Merge #4393: M2: an Indexed image with a short lookup table or an out-of-range hival is drawn as most readers draw it, and one with no table is skipped with a warning (view.render-pages)

# Conflicts:
#	vendor/README.md

apps/pdfcraft-cli/src/commands.rs

@@ -440,6 +440,12 @@ pub(crate) fn render(args: &[String]) -> Result<(), CliError> {
                 let _ =
                     writeln!(std::io::stderr().lock(), "warning: page {page}: a colour that could not be computed was painted in a fallback colour");
             }
+            pdfcraft_render::RenderWarning::MalformedIndexedImage => {
+                let _ = writeln!(
+                    std::io::stderr().lock(),
+                    "warning: page {page}: an image with a malformed Indexed colour table may be drawn wrongly or not at all"
+                );
+            }
             pdfcraft_render::RenderWarning::Type3WidthsMissing => {
                 let _ = writeln!(
                     std::io::stderr().lock(),

crates/automation/src/pages.rs

@@ -414,6 +414,7 @@ fn render_warning_text(warning: pdfcraft_render::RenderWarning) -> &'static str
     match warning {
         RenderWarning::ContentTruncated => "content past the page's safety budget was skipped",
         RenderWarning::ColourFallback => "a colour that could not be computed was painted in a fallback colour",
+        RenderWarning::MalformedIndexedImage => "an image with a malformed Indexed colour table may be drawn wrongly or not at all",
         RenderWarning::Type3WidthsMissing => "a Type 3 font has no /Widths; its glyphs were spaced by the widths their glyph procedures give",
     }
 }

crates/automation/tests/it/render_warnings.rs

@@ -96,6 +96,33 @@ fn a_clean_page_renders_to_the_png_alone() {
     let _ = std::fs::remove_dir_all(dir);
 }
 
+/// An image whose Indexed lookup table is too short (ISO 32000-2 §8.6.6.3) is drawn recovered, and
+/// `page_render` says so in the words `pdfcraft-cli render` prints.
+#[test]
+fn page_render_reports_a_recovered_malformed_indexed_image() {
+    let dir = std::env::temp_dir().join(format!("pdfcraft-render-warnings-indexed-{}", std::process::id()));
+    let _ = std::fs::remove_dir_all(&dir);
+    std::fs::create_dir_all(&dir).unwrap();
+    let content = "q 200 0 0 200 0 0 cm /Im Do Q";
+    // hival 7 over DeviceRGB needs 24 bytes; the table holds 6.
+    let pdf = format!(
+        "%PDF-1.7\n1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj\n\
+         2 0 obj << /Type /Pages /Kids [3 0 R] /Count 1 >> endobj\n\
+         3 0 obj << /Type /Page /Parent 2 0 R /MediaBox [0 0 200 200] /Resources << /XObject << /Im 5 0 R >> >> /Contents 4 0 R >> endobj\n\
+         4 0 obj << /Length {} >> stream\n{content}\nendstream endobj\n\
+         5 0 obj << /Type /XObject /Subtype /Image /Width 2 /Height 1 /BitsPerComponent 8 /ColorSpace [/Indexed /DeviceRGB 7 <FF0000 00FF00>] /Filter /ASCIIHexDecode /Length 5 >> stream\n0001>\nendstream endobj\n\
+         trailer << /Root 1 0 R >>\n%%EOF",
+        content.len()
+    );
+    std::fs::write(dir.join("indexed.pdf"), pdf).unwrap();
+    let mut a = Automation::new().with_root(&dir).unwrap();
+    let doc = open(&mut a, "indexed.pdf");
+    let out = a.call("page_render", &json!({ "doc": doc, "page": 1, "dpi": 72 })).unwrap();
+    let Some(Content::Json(report)) = out.get(1) else { panic!("expected the warnings as JSON: {out:?}") };
+    assert_eq!(report["warnings"], json!(["an image with a malformed Indexed colour table may be drawn wrongly or not at all"]));
+    let _ = std::fs::remove_dir_all(dir);
+}
+
 /// Over MCP the warnings are a text block after the image and the call's `structuredContent`.
 #[cfg(feature = "mcp")]
 #[test]

crates/render/src/raster.rs

@@ -270,6 +270,13 @@ pub enum RenderWarning {
     /// A colour could not be computed (a colour space that cannot be read, or a Separation or
     /// DeviceN tint transform that cannot be evaluated) and was painted in a fallback colour.
     ColourFallback,
+    /// An image's `/Indexed` colour space is malformed (ISO 32000-2 §8.6.6.3), so the image may be
+    /// drawn wrongly or not at all. It is drawn recovered when it can be: indices past a lookup
+    /// table shorter than m × (hival + 1) bytes in black, a `hival` past 255 taken as 255, a real
+    /// one truncated, or (a `hival` below zero or not a number) the image in DeviceGray. With no
+    /// lookup table at all (missing, not a string or a stream, or a stream that does not decode)
+    /// it is not drawn. Reported once per page however many images it affects.
+    MalformedIndexedImage,
     /// A Type 3 font without the `/Widths` it requires (ISO 32000-2 Table 110) was used; its
     /// glyphs were spaced by the widths their glyph procedures give.
     Type3WidthsMissing,
@@ -477,10 +484,12 @@ fn warning_settings(settings: &InterpreterSettings, warnings: Arc<Mutex<Vec<Rend
     InterpreterSettings {
         warning_sink: Arc::new(move |warning| {
             // Nested forms, Type 3 glyphs and patterns are interpreted again and each reports the
-            // exhausted budget, or the colours that fell back: once per page is enough.
+            // exhausted budget, or the colours that fell back, and several images can share one
+            // malformed colour space: each warning is recorded once per page.
             let warning = match warning {
                 InterpreterWarning::ContentTruncated => RenderWarning::ContentTruncated,
                 InterpreterWarning::ColorFallback => RenderWarning::ColourFallback,
+                InterpreterWarning::MalformedIndexedImage => RenderWarning::MalformedIndexedImage,
                 InterpreterWarning::Type3WidthsMissing => RenderWarning::Type3WidthsMissing,
                 InterpreterWarning::UnsupportedFont | InterpreterWarning::ImageDecodeFailure => return,
             };

crates/render/src/raster/tests.rs

@@ -9282,6 +9282,10 @@ fn an_averaged_down_soft_mask_follows_its_samples_and_its_drawn_size() {
 // ICC profiles whose stream dictionaries are the same (vendored hayro-interpret patch 48).
 mod icc_profile_cache;
 
+// An Indexed resource named by many inline images reads its lookup once (vendored
+// hayro-interpret patch 80).
+mod indexed_lookup_cache;
+
 // A Type 3 glyph shown again decodes its procedure once (vendored hayro-interpret patch 87).
 mod type3_proc_cache;
 

crates/render/src/raster/tests/indexed_lookup_cache.rs

@@ -0,0 +1,75 @@
+//! An `/Indexed` colour space named by a resource is built once, and its lookup table read
+//! once, however many inline images name it (vendored hayro-interpret patch 80). An earlier
+//! version of the patch checked the table again before every inline image, decoding a lookup
+//! stream once per image outside the colour-space cache.
+
+use super::image_cache_pdf;
+
+/// A page that draws `count` 2 × 1 inline images in the resource colour space `/P`, an Indexed
+/// space over DeviceRGB with `hival` whose lookup is stream object 5 holding `table`.
+fn page(count: usize, hival: u8, table: &[u8]) -> Vec<u8> {
+    let image = "q 20 0 0 20 0 0 cm BI /W 2 /H 1 /BPC 8 /CS /P /F /AHx ID 0001> EI Q";
+    let content = vec![image; count].join("\n");
+    let resources = format!("<< /ColorSpace << /P [/Indexed /DeviceRGB {hival} 5 0 R] >> >>");
+    let dict = format!("<< /Length {} >>", table.len());
+    image_cache_pdf(&content, &resources, &[(&dict, table)])
+}
+
+fn lookup_reads(pdf: Vec<u8>) -> [usize; 2] {
+    use hayro::hayro_interpret::font::Glyph;
+    use hayro::hayro_interpret::hayro_syntax::Pdf;
+    use hayro::hayro_interpret::{
+        BlendMode, ClipPath, Context, Device, GlyphDrawMode, Image, InterpreterCache, InterpreterSettings, Paint, PathDrawMode, SoftMask,
+        TransformExt, interpret_page,
+    };
+    use kurbo::{Affine, BezPath, Rect};
+    use std::sync::Arc;
+
+    /// Draws nothing: only the colour spaces the page reads matter here.
+    struct Nothing;
+    impl<'a> Device<'a> for Nothing {
+        fn set_soft_mask(&mut self, _: Option<SoftMask<'a>>) {}
+        fn set_blend_mode(&mut self, _: BlendMode) {}
+        fn draw_path(&mut self, _: &BezPath, _: Affine, _: &Paint<'a>, _: &PathDrawMode) {}
+        fn push_clip_path(&mut self, _: &ClipPath) {}
+        fn push_transparency_group(&mut self, _: f32, _: Option<SoftMask<'a>>, _: BlendMode) {}
+        fn draw_glyph(&mut self, _: &Glyph<'a>, _: Affine, _: Affine, _: &Paint<'a>, _: &GlyphDrawMode) {}
+        fn draw_image(&mut self, _: Image<'a, '_>, _: Affine) {}
+        fn pop_clip_path(&mut self) {}
+        fn pop_transparency_group(&mut self) {}
+    }
+
+    let pdf = Pdf::new(Arc::new(pdf)).expect("parses");
+    let page = &pdf.pages()[0];
+    let cache = InterpreterCache::new();
+    [(); 2].map(|()| {
+        let initial = page.initial_transform(true).to_kurbo();
+        let mut ctx = Context::new(initial, Rect::new(0.0, 0.0, 130.0, 70.0), &cache, page.xref(), InterpreterSettings::default());
+        interpret_page(page, &mut ctx, &mut Nothing);
+        cache.indexed_lookup_reads()
+    })
+}
+
+/// Twenty inline images naming one resource read its lookup stream once, in the first render
+/// and none in a second render through the same cache: for a full table, and for a short one
+/// that is padded (whose images an earlier version of the patch checked, and skipped, one by
+/// one).
+#[test]
+fn inline_images_naming_one_indexed_resource_read_its_lookup_once() {
+    let full = page(20, 1, b"\xFF\x00\x00\x00\xFF\x00");
+    assert_eq!(lookup_reads(full), [1, 1], "a full table: read once, then cached");
+    let short = page(20, 7, b"\xFF\x00\x00\x00\xFF\x00");
+    assert_eq!(lookup_reads(short), [1, 1], "a short table: read once, then cached");
+}
+
+/// Two Indexed resources whose arrays differ (`/P` names object 5, `/Q` object 6) are two
+/// spaces: each is built, and its table read, once, and neither stands in for the other.
+#[test]
+fn two_indexed_resources_are_cached_apart() {
+    let image = |name: &str| format!("q 20 0 0 20 0 0 cm BI /W 2 /H 1 /BPC 8 /CS /{name} /F /AHx ID 0001> EI Q");
+    let content = [image("P"), image("Q"), image("P"), image("Q")].join("\n");
+    let resources = "<< /ColorSpace << /P [/Indexed /DeviceRGB 1 5 0 R] /Q [/Indexed /DeviceRGB 1 6 0 R] >> >>";
+    let (red_green, blue_white): (&[u8], &[u8]) = (b"\xFF\x00\x00\x00\xFF\x00", b"\x00\x00\xFF\xFF\xFF\xFF");
+    let pdf = image_cache_pdf(&content, resources, &[("<< /Length 6 >>", red_green), ("<< /Length 6 >>", blue_white)]);
+    assert_eq!(lookup_reads(pdf), [2, 2], "each table read once");
+}

crates/render/tests/it/indexed_lookup.rs

@@ -0,0 +1,458 @@
+//! An `/Indexed` image whose colour space breaks ISO 32000-2 §8.6.6.3 (a lookup table shorter
+//! than m × (hival + 1) bytes, or a `hival` past 255 or with a fraction) is drawn as most
+//! readers draw it, and the render reports `RenderWarning::MalformedIndexedImage`:
+//! - a short table keeps the entries it holds in full, and an index past them is drawn black,
+//!   whatever the base space;
+//! - a `hival` past 255 is taken as 255;
+//! - a real `hival` is truncated toward zero.
+//!
+//! An image whose lookup is missing, `null`, neither a string nor a stream, or a stream that
+//! does not decode is not drawn, and is reported. A `hival` that is negative or not a number
+//! keeps the upstream handling (the image falls back to DeviceGray), and is reported too.
+//! Upstream fell back to DeviceGray for every one of these, painting the indices as near-black
+//! grey levels, even those whose entries were present.
+//!
+//! The images are 8 × 1, samples 0 to 7, each column 25 px wide on a 200 × 200 page (the layout
+//! of the seven cases A to G a reviewer compared across readers on #4393).
+
+use std::sync::Arc;
+
+use pdfcraft_render::{PageRenderer, RenderConfig, RenderRequest, RenderWarning, RenderedPage, RequestKind};
+
+/// The eight entries the tests' tables hold, in order.
+const TABLE: &str = "FF0000 00FF00 0000FF FFFF00 FF00FF 00FFFF 808080 000000";
+const COLOURS: [[u8; 3]; 8] = [[255, 0, 0], [0, 255, 0], [0, 0, 255], [255, 255, 0], [255, 0, 255], [0, 255, 255], [128, 128, 128], [0, 0, 0]];
+const BLACK: [u8; 3] = [0, 0, 0];
+const WHITE: [u8; 3] = [255, 255, 255];
+const DRAW_IMAGE: &str = "q 200 0 0 200 0 0 cm /Im Do Q";
+const MALFORMED: [RenderWarning; 1] = [RenderWarning::MalformedIndexedImage];
+
+/// The first `n` hex entries of `TABLE`.
+fn entries(n: usize) -> String {
+    TABLE.split(' ').take(n).collect::<Vec<_>>().join(" ")
+}
+
+/// A 200 × 200 page whose `/Im` image is eight samples wide (indices 0 to 7) and covers the
+/// page, with `/P` naming `colour_space` as a resource. `extra` objects are numbered from 6.
+fn pdf(colour_space: &str, content: &str, extra: &[&[u8]]) -> Vec<u8> {
+    pdf_with_resource("P", colour_space, content, extra)
+}
+
+/// As `pdf`, with the colour space resource named `resource` rather than `/P`.
+fn pdf_with_resource(resource: &str, colour_space: &str, content: &str, extra: &[&[u8]]) -> Vec<u8> {
+    pdf_with_resources(resource, colour_space, "", content, extra)
+}
+
+/// As `pdf_with_resource`, with `more` entries in the page's resource dictionary.
+fn pdf_with_resources(resource: &str, colour_space: &str, more: &str, content: &str, extra: &[&[u8]]) -> Vec<u8> {
+    let mut objs: Vec<Vec<u8>> = vec![
+        b"<< /Type /Catalog /Pages 2 0 R >>".to_vec(),
+        b"<< /Type /Pages /Kids [3 0 R] /Count 1 >>".to_vec(),
+        format!(
+            "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 200 200] /Contents 4 0 R \
+             /Resources << /XObject << /Im 5 0 R >> /ColorSpace << /{resource} {colour_space} >> {more} >> >>"
+        )
+        .into_bytes(),
+        format!("<< /Length {} >>\nstream\n{content}\nendstream", content.len()).into_bytes(),
+        format!(
+            "<< /Type /XObject /Subtype /Image /Width 8 /Height 1 /BitsPerComponent 8 \
+             /ColorSpace {colour_space} /Filter /ASCIIHexDecode /Length 17 >>\nstream\n0001020304050607>\nendstream"
+        )
+        .into_bytes(),
+    ];
+    objs.extend(extra.iter().map(|o| o.to_vec()));
+    let mut out = b"%PDF-1.7\n".to_vec();
+    let mut offsets = Vec::new();
+    for (i, body) in objs.iter().enumerate() {
+        offsets.push(out.len());
+        out.extend_from_slice(format!("{} 0 obj\n", i + 1).as_bytes());
+        out.extend_from_slice(body);
+        out.extend_from_slice(b"\nendobj\n");
+    }
+    let xref = out.len();
+    let size = objs.len() + 1;
+    out.extend_from_slice(format!("xref\n0 {size}\n0000000000 65535 f \n").as_bytes());
+    for off in offsets {
+        out.extend_from_slice(format!("{off:010} 00000 n \n").as_bytes());
+    }
+    out.extend_from_slice(format!("trailer\n<< /Size {size} /Root 1 0 R >>\nstartxref\n{xref}\n%%EOF\n").as_bytes());
+    out
+}
+
+fn render(bytes: Vec<u8>) -> RenderedPage {
+    let page = PageRenderer::new(Arc::new(bytes), RenderConfig::default()).render(RenderRequest {
+        page: 0,
+        kind: RequestKind::Pixels,
+        tile: None,
+        scale: 1.0,
+        tag: 0,
+        preview: false,
+    });
+    assert!(page.error.is_none(), "{:?}", page.error);
+    assert_eq!((page.width, page.height), (200, 200));
+    page
+}
+
+/// The centre pixel of each of the eight columns (indices 0 to 7), RGB.
+fn columns(page: &RenderedPage) -> [[u8; 3]; 8] {
+    std::array::from_fn(|i| {
+        let at = (100 * page.width as usize + 12 + 25 * i) * 4;
+        let px = page.rgba.get(at..at + 3).expect("pixel inside the page");
+        [px[0], px[1], px[2]]
+    })
+}
+
+/// The first `n` of `COLOURS`, then black: a table of `n` entries.
+fn padded(n: usize) -> [[u8; 3]; 8] {
+    std::array::from_fn(|i| if i < n { COLOURS[i] } else { BLACK })
+}
+
+/// Index `i` drawn as the grey level `i` / 255: the DeviceGray fallback.
+fn grey_fallback() -> [[u8; 3]; 8] {
+    std::array::from_fn(|i| [i as u8; 3])
+}
+
+#[track_caller]
+fn assert_page(colour_space: &str, extra: &[&[u8]], want: [[u8; 3]; 8], warned: bool) {
+    let page = render(pdf(colour_space, DRAW_IMAGE, extra));
+    assert_eq!(columns(&page), want, "{colour_space}: the columns");
+    let want_warnings: &[RenderWarning] = if warned { &MALFORMED } else { &[] };
+    assert_eq!(page.warnings, want_warnings, "{colour_space}: the warnings");
+}
+
+/// Case A: a full table draws its eight colours, with no warning.
+#[test]
+fn a_full_table_draws_its_colours() {
+    assert_page(&format!("[/Indexed /DeviceRGB 7 <{TABLE}>]"), &[], COLOURS, false);
+    assert_page(&format!("[/Indexed /DeviceRGB 7.0 <{TABLE}>]"), &[], COLOURS, false);
+}
+
+/// Cases B and F: a short table keeps the entries it holds, and an index past them is black.
+#[test]
+fn a_short_table_draws_its_entries_and_black_past_them() {
+    // B: hival 7 needs 24 bytes; 6 are given. Indices 0 and 1 used to draw black.
+    assert_page(&format!("[/Indexed /DeviceRGB 7 <{}>]", entries(2)), &[], padded(2), true);
+    // F: one entry short.
+    assert_page(&format!("[/Indexed /DeviceRGB 7 <{}>]", entries(7)), &[], padded(7), true);
+    // One byte short: entry 1 is incomplete, so it is not held; indices past hival take it.
+    let mut want = [BLACK; 8];
+    want[0] = COLOURS[0];
+    assert_page("[/Indexed /DeviceRGB 1 <FF0000 00FF>]", &[], want, true);
+}
+
+/// The base space for each of the short-table tests below, with its two-entry table, and the
+/// stream objects it needs (numbered from 6).
+fn short_table_in(base: &str, table: &str) -> String {
+    format!("[/Indexed {base} 7 <{table}>]")
+}
+
+/// A short table over `base`: the two entries it holds draw as they do with a full table (the
+/// rest of which is white, never drawn black by a reader), and every index past them is black.
+/// Zero bytes would be white in DeviceCMYK, Separation and DeviceN, and blue in Lab.
+#[track_caller]
+fn assert_black_past_the_table(base: &str, two: &str, white: &str, extra: &[&[u8]]) {
+    let full = render(pdf(&short_table_in(base, &format!("{two} {}", white.repeat(6))), DRAW_IMAGE, extra));
+    assert!(full.warnings.is_empty(), "{base}: {:?}", full.warnings);
+    let full = columns(&full);
+    assert!(full[..2].iter().all(|c| *c != BLACK && *c != WHITE), "{base}: the two entries have colour: {full:?}");
+    let page = render(pdf(&short_table_in(base, two), DRAW_IMAGE, extra));
+    assert_eq!(page.warnings, MALFORMED, "{base}");
+    let mut want = [BLACK; 8];
+    want[..2].copy_from_slice(&full[..2]);
+    assert_eq!(columns(&page), want, "{base}: entries 2 to 7 are black");
+}
+
+/// DeviceGray: zero is black here, so a zero-padded table drew the same (a guard).
+#[test]
+fn indices_past_a_short_table_are_black_over_device_gray() {
+    assert_black_past_the_table("/DeviceGray", "40 80", "FF ", &[]);
+}
+
+/// DeviceCMYK: a zero entry would be white (no ink).
+#[test]
+fn indices_past_a_short_table_are_black_over_device_cmyk() {
+    assert_black_past_the_table("/DeviceCMYK", "FF000000 00FF0000", "00000000 ", &[]);
+}
+
+/// Lab: a zero entry would be L* 0 with a* and b* at the bottom of their ranges, blue.
+#[test]
+fn indices_past_a_short_table_are_black_over_lab() {
+    assert_black_past_the_table("[/Lab << /WhitePoint [0.9505 1 1.089] /Range [-128 127 -128 127] >>]", "80FF80 8080FF", "FF8080 ", &[]);
+}
+
+/// Separation (a spot orange: tint 1 is orange in DeviceCMYK): a zero entry would be tint 0,
+/// white.
+#[test]
+fn indices_past_a_short_table_are_black_over_separation() {
+    assert_black_past_the_table(
+        "[/Separation /Orange /DeviceCMYK << /FunctionType 2 /Domain [0 1] /C0 [0 0 0 0] /C1 [0 0.5 1 0] /N 1 >>]",
+        "FF 80",
+        "00 ",
+        &[],
+    );
+}
+
+/// DeviceN (cyan and magenta, mapped onto DeviceCMYK by a Type 4 function): a zero entry would
+/// be white.
+#[test]
+fn indices_past_a_short_table_are_black_over_device_n() {
+    let tint: &[u8] = b"<< /FunctionType 4 /Domain [0 1 0 1] /Range [0 1 0 1 0 1 0 1] /Length 9 >>\nstream\n{ 0 0 }\nendstream";
+    assert_black_past_the_table("[/DeviceN [/Cyan /Magenta] /DeviceCMYK 6 0 R]", "FF00 00FF", "0000 ", &[tint]);
+}
+
+/// ICCBased with four components and no usable profile, drawn in its /Alternate DeviceCMYK: a
+/// zero entry would be white.
+#[test]
+fn indices_past_a_short_table_are_black_over_icc_based_with_four_components() {
+    let profile: &[u8] = b"<< /N 4 /Alternate /DeviceCMYK /Length 4 >>\nstream\nJUNK\nendstream";
+    assert_black_past_the_table("[/ICCBased 6 0 R]", "FF000000 00FF0000", "00000000 ", &[profile]);
+}
+
+/// Case C: an index past `hival` takes the nearest entry (ISO 32000-2 §8.6.6.3: "shall be
+/// adjusted to the nearest value within that range"); a table longer than needed is fine.
+#[test]
+fn an_index_past_hival_takes_the_last_entry() {
+    let mut want = COLOURS;
+    want[4..].fill(COLOURS[3]);
+    assert_page(&format!("[/Indexed /DeviceRGB 3 <{TABLE}>]"), &[], want, false);
+}
+
+/// Case D: a `hival` past 255 is taken as 255, and the table padded to 256 entries.
+#[test]
+fn a_hival_past_255_is_clamped() {
+    assert_page(&format!("[/Indexed /DeviceRGB 300 <{TABLE}>]"), &[], COLOURS, true);
+    // With a table long enough for 301 entries, the clamp alone is reported.
+    let long = format!("<{TABLE}{}>", "00".repeat(3 * 293));
+    assert_page(&format!("[/Indexed /DeviceRGB 300 {long}]"), &[], COLOURS, true);
+    // 256, one past.
+    assert_page(&format!("[/Indexed /DeviceRGB 256 <{TABLE}{}>]", "00".repeat(3 * 249)), &[], COLOURS, true);
+    // 255, the largest, with a full table draws without a warning.
+    assert_page(&format!("[/Indexed /DeviceRGB 255 <{TABLE}{}>]", "00".repeat(3 * 248)), &[], COLOURS, false);
+}
+
+/// Case E: a real `hival` is truncated toward zero, as hayro read it before patch (80).
+#[test]
+fn a_real_hival_is_truncated() {
+    assert_page(&format!("[/Indexed /DeviceRGB 7.5 <{TABLE}>]"), &[], COLOURS, true);
+    // 6.9 is 6, not 7: index 7 takes entry 6.
+    let mut want = COLOURS;
+    want[7] = COLOURS[6];
+    assert_page(&format!("[/Indexed /DeviceRGB 6.9 <{TABLE}>]"), &[], want, true);
+    // -0.5 is 0 (toward zero; a real above -1 truncates to 0): every index takes entry 0.
+    assert_page(&format!("[/Indexed /DeviceRGB -0.5 <{TABLE}>]"), &[], [COLOURS[0]; 8], true);
+    // 255.5 is 255; 255.0 written as a real is the integer 255.
+    let full = format!("<{TABLE}{}>", "00".repeat(3 * 248));
+    assert_page(&format!("[/Indexed /DeviceRGB 255.5 {full}]"), &[], COLOURS, true);
+    assert_page(&format!("[/Indexed /DeviceRGB 255.0 {full}]"), &[], COLOURS, false);
+}
+
+/// Case G: inline images, named through the resources or given in place, are recovered too, and
+/// skipped when the space has no table.
+#[test]
+fn inline_images_with_a_short_table_are_recovered_too() {
+    let inline = |cs: &str| format!("q 200 0 0 200 0 0 cm BI /W 8 /H 1 /BPC 8 /CS {cs} /F /AHx ID 0001020304050607> EI Q");
+    let short = format!("[/Indexed /DeviceRGB 7 <{}>]", entries(2));
+    // Named through the page's resources.
+    let page = render(pdf_with_resource("CS0", &short, &inline("/CS0"), &[]));
+    assert_eq!(columns(&page), padded(2));
+    assert_eq!(page.warnings, MALFORMED);
+    // Given in place, with the inline abbreviations.
+    let page = render(pdf("/DeviceRGB", &inline(&format!("[/I /RGB 7 <{}>]", entries(2))), &[]));
+    assert_eq!(columns(&page), padded(2));
+    assert_eq!(page.warnings, MALFORMED);
+    // A `hival` with no reading falls back and is reported, named or in place.
+    let page = render(pdf(&format!("[/Indexed /DeviceRGB -1 <{TABLE}>]"), &inline("/P"), &[]));
+    assert_eq!((columns(&page), page.warnings), (grey_fallback(), MALFORMED.to_vec()));
+    let page = render(pdf("/DeviceRGB", &inline(&format!("[/I /RGB -1 <{TABLE}>]")), &[]));
+    assert_eq!((columns(&page), page.warnings), (grey_fallback(), MALFORMED.to_vec()));
+    // No table: not drawn, named or in place.
+    let page = render(pdf("[/Indexed /DeviceRGB 7 null]", &inline("/P"), &[]));
+    assert_eq!((columns(&page), page.warnings), ([WHITE; 8], MALFORMED.to_vec()));
+    let page = render(pdf("/DeviceRGB", &inline("[/I /RGB 7]"), &[]));
+    assert_eq!((columns(&page), page.warnings), ([WHITE; 8], MALFORMED.to_vec()));
+    // A well-formed one draws without a warning.
+    let page = render(pdf(&format!("[/Indexed /DeviceRGB 7 <{TABLE}>]"), &inline("/P"), &[]));
+    assert_eq!(columns(&page), COLOURS);
+    assert!(page.warnings.is_empty(), "{:?}", page.warnings);
+}
+
+#[test]
+fn an_inline_device_space_name_ignores_a_resource_of_the_same_name() {
+    // ISO 32000-2 §8.6.8 (the CS operator): the device colour space names never refer to
+    // resources in the ColorSpace subdictionary, so a malformed Indexed resource that shares the
+    // name (here the inline abbreviation `/RGB`, and `/DeviceRGB`) is not used.
+    let data: String = TABLE.split(' ').collect();
+    for (resource, cs) in [("RGB", "/RGB"), ("DeviceRGB", "/DeviceRGB")] {
+        let content = format!("q 200 0 0 200 0 0 cm BI /W 8 /H 1 /BPC 8 /CS {cs} /F /AHx ID {data}> EI Q");
+        let page = render(pdf_with_resource(resource, "[/Indexed /DeviceRGB 7 <FF0000 00FF00>]", &content, &[]));
+        assert_eq!(columns(&page), COLOURS, "{cs}: the device space draws");
+        assert!(page.warnings.is_empty(), "{cs}: {:?}", page.warnings);
+    }
+}
+
+#[test]
+fn a_page_with_several_malformed_images_reports_the_warning_once() {
+    let content = format!("{DRAW_IMAGE} {DRAW_IMAGE} 0 0 1 rg 0 0 10 10 re f");
+    let page = render(pdf(&format!("[/Indexed /DeviceRGB 7 <{}>]", entries(2)), &content, &[]));
+    assert_eq!(page.warnings, MALFORMED);
+    assert_eq!(columns(&page), padded(2));
+    let at = (195 * page.width as usize + 5) * 4;
+    assert_eq!(page.rgba.get(at..at + 3), Some(&[0u8, 0, 255][..]), "content after the images still draws");
+}
+
+#[test]
+fn a_lookup_string_given_by_reference_is_read() {
+    // A lookup is a byte string or a stream (ISO 32000-2 §8.6.6.3), and either may be indirect.
+    // A string by reference used to be missed: trying it as a stream consumed the reference.
+    let full = format!("<{TABLE}>");
+    assert_page("[/Indexed /DeviceRGB 7 6 0 R]", &[full.as_bytes()], COLOURS, false);
+    let short = format!("<{}>", entries(2));
+    assert_page("[/Indexed /DeviceRGB 7 6 0 R]", &[short.as_bytes()], padded(2), true);
+    // Something that is neither a string nor a stream is no table: the image is not drawn.
+    assert_page("[/Indexed /DeviceRGB 7 6 0 R]", &[b"42"], [WHITE; 8], true);
+}
+
+#[test]
+fn the_table_size_holds_at_the_bounds_of_hival_and_the_base() {
+    // Four components (DeviceCMYK): hival 1 needs 8 bytes. With 7, entry 1 is incomplete, so it
+    // is not held: index 1, and the indices past hival that take it, are black.
+    let full = render(pdf("[/Indexed /DeviceCMYK 1 <FF000000 00FF0000>]", DRAW_IMAGE, &[]));
+    assert!(full.warnings.is_empty(), "{:?}", full.warnings);
+    assert_ne!(columns(&full)[0], WHITE, "the image is drawn");
+    let short = render(pdf("[/Indexed /DeviceCMYK 1 <FF000000 00FF00>]", DRAW_IMAGE, &[]));
+    assert_eq!(short.warnings, MALFORMED);
+    let mut want = [BLACK; 8];
+    want[0] = columns(&full)[0];
+    assert_eq!(columns(&short), want);
+    // One component (DeviceGray): hival 0 needs one byte; an empty table holds no entry.
+    assert_page("[/Indexed /DeviceGray 0 <80>]", &[], [[128; 3]; 8], false);
+    assert_page("[/Indexed /DeviceGray 0 <>]", &[], [BLACK; 8], true);
+    // hival 255 over DeviceRGB needs 768 bytes: 767 is one short.
+    assert_page(&format!("[/Indexed /DeviceRGB 255 <{TABLE}{}>]", "00".repeat(3 * 248 - 1)), &[], COLOURS, true);
+}
+
+#[test]
+fn a_hival_that_would_overflow_a_size_is_clamped_without_panicking() {
+    // Values whose (hival + 1) × components would overflow a 32-bit or 64-bit size, or that do
+    // not fit an integer at all, are taken as 255.
+    for hival in ["4294967295", "4294967296", "18446744073709551615", "99999999999999999999999"] {
+        assert_page(&format!("[/Indexed /DeviceRGB {hival} <{}>]", entries(2)), &[], padded(2), true);
+    }
+}
+
+/// A `hival` that is negative or not a number has no reading to recover: the image keeps the
+/// upstream handling (it falls back to DeviceGray), and the warning is still reported.
+#[test]
+fn a_negative_or_non_numeric_hival_falls_back_with_a_warning() {
+    for hival in ["-1", "-1.5", "-2147483649", "/One", "null"] {
+        assert_page(&format!("[/Indexed /DeviceRGB {hival} <{TABLE}>]"), &[], grey_fallback(), true);
+    }
+}
+
+/// An empty string is a table that holds no entry: every index is black.
+#[test]
+fn an_empty_table_draws_black() {
+    for lookup in ["<>", "()"] {
+        assert_page(&format!("[/Indexed /DeviceRGB 1 {lookup}]"), &[], padded(0), true);
+    }
+}
+
+/// A lookup that is missing, `null` or a number is no table at all: the image is not drawn (the
+/// page stays white), and the warning is reported.
+#[test]
+fn a_missing_lookup_skips_the_image() {
+    for lookup in ["", "null", "42"] {
+        assert_page(&format!("[/Indexed /DeviceRGB 7 {lookup}]"), &[], [WHITE; 8], true);
+    }
+}
+
+#[test]
+fn a_lookup_stream_is_measured_after_decoding() {
+    let short: &[u8] = b"<< /Length 14 /Filter /ASCIIHexDecode >>\nstream\nFF0000 00FF00>\nendstream";
+    assert_page("[/Indexed /DeviceRGB 7 6 0 R]", &[short], padded(2), true);
+    let mut want = [COLOURS[1]; 8];
+    want[0] = COLOURS[0];
+    assert_page("[/Indexed /DeviceRGB 1 6 0 R]", &[short], want, false);
+    // An empty stream decodes to an empty table, which holds no entry.
+    let empty: &[u8] = b"<< /Length 0 >>\nstream\n\nendstream";
+    assert_page("[/Indexed /DeviceRGB 1 6 0 R]", &[empty], padded(0), true);
+}
+
+/// A stream that does not decode is no table: the image is not drawn, and is reported.
+#[test]
+fn a_lookup_stream_that_cannot_be_decoded_skips_the_image() {
+    for broken in [
+        &b"<< /Length 12 /Filter /FlateDecode >>\nstream\nnot deflated\nendstream"[..],
+        b"<< /Length 6 /Filter /FlateDecode >>\nstream\nnot-zl\nendstream",
+    ] {
+        assert_page("[/Indexed /DeviceRGB 7 6 0 R]", &[broken], [WHITE; 8], true);
+    }
+}
+
+/// An Indexed space whose base cannot be read keeps the upstream handling (DeviceGray), whatever
+/// its `hival`; only a `hival` with no reading is reported.
+#[test]
+fn an_unreadable_base_keeps_its_fallback() {
+    for (hival, warned) in [("1", false), ("300", false), ("-1", true)] {
+        assert_page(&format!("[/Indexed /NoSuchSpace {hival} <{TABLE}>]"), &[], grey_fallback(), warned);
+    }
+}
+
+/// A fill or stroke colour set in such a space (`cs`, `sc`, `CS`, `SC`) draws through the same
+/// recovered table, and the recovery is reported as `ColourFallback`, as the upstream DeviceGray
+/// fallback was.
+#[test]
+fn a_fill_or_stroke_colour_in_a_short_table_is_recovered_and_reported() {
+    let fill = |index: u8| format!("/P cs {index} sc 0 0 200 200 re f");
+    let short = format!("[/Indexed /DeviceRGB 7 <{}>]", entries(2));
+    let page = render(pdf(&short, &fill(1), &[]));
+    assert_eq!(columns(&page)[0], COLOURS[1], "entry 1 is green");
+    assert_eq!(page.warnings, vec![RenderWarning::ColourFallback]);
+    let page = render(pdf(&short, &fill(5), &[]));
+    assert_eq!(columns(&page)[0], BLACK, "entry 5 is past the table");
+    // Over DeviceCMYK too, where a zero entry would be white.
+    let cmyk = "[/Indexed /DeviceCMYK 7 <FF000000 00FF0000>]";
+    let page = render(pdf(cmyk, &fill(5), &[]));
+    assert_eq!(columns(&page)[0], BLACK, "DeviceCMYK entry 5 is past the table");
+    // A stroke colour (`CS`, `SC`) the same way.
+    let page = render(pdf(&short, "/P CS 1 SC 20 w 0 100 m 200 100 l S", &[]));
+    assert_eq!(columns(&page)[0], COLOURS[1], "the stroke is green");
+    assert_eq!(page.warnings, vec![RenderWarning::ColourFallback]);
+    let page = render(pdf(&format!("[/Indexed /DeviceRGB 7 <{TABLE}>]"), &fill(1), &[]));
+    assert_eq!(columns(&page)[0], COLOURS[1]);
+    assert!(page.warnings.is_empty(), "{:?}", page.warnings);
+}
+
+/// A fill colour in an Indexed space with no lookup table keeps the upstream fallback: DeviceGray,
+/// reported as `ColourFallback` (tint 1 is white, over a red fill).
+#[test]
+fn a_fill_colour_in_a_space_with_no_table_keeps_the_fallback() {
+    for lookup in ["", "null"] {
+        let content = "1 0 0 rg 0 0 200 200 re f /P cs 1 sc 0 0 200 200 re f";
+        let page = render(pdf(&format!("[/Indexed /DeviceRGB 7 {lookup}]"), content, &[]));
+        assert_eq!(columns(&page)[0], WHITE, "{lookup:?}: DeviceGray 1");
+        assert_eq!(page.warnings, vec![RenderWarning::ColourFallback], "{lookup:?}");
+        // A stroke colour the same way.
+        let content = "1 0 0 rg 0 0 200 200 re f /P CS 1 SC 20 w 0 100 m 200 100 l S";
+        let page = render(pdf(&format!("[/Indexed /DeviceRGB 7 {lookup}]"), content, &[]));
+        assert_eq!(columns(&page)[0], WHITE, "{lookup:?}: a DeviceGray 1 stroke");
+        assert_eq!(page.warnings, vec![RenderWarning::ColourFallback], "{lookup:?}");
+    }
+}
+
+/// Only images and the colour operators handle a space with no table: anything else reads it as
+/// upstream did, as a space that cannot be read. A shading in one is not drawn.
+#[test]
+fn a_shading_in_a_space_with_no_table_is_not_drawn() {
+    let shading = |cs: &str| {
+        format!(
+            "/Shading << /Sh << /ShadingType 2 /ColorSpace {cs} /Coords [0 0 200 0] \
+             /Function << /FunctionType 2 /Domain [0 1] /C0 [0] /C1 [1] /N 1 >> /Extend [true true] >> >>"
+        )
+    };
+    let page = render(pdf_with_resources("P", "/DeviceRGB", &shading("[/Indexed /DeviceRGB 1 null]"), "/Sh sh", &[]));
+    assert_eq!(columns(&page), [WHITE; 8]);
+    // The same shading with a table draws.
+    let page = render(pdf_with_resources("P", "/DeviceRGB", &shading("[/Indexed /DeviceRGB 1 <FF0000 00FF00>]"), "/Sh sh", &[]));
+    assert_ne!(columns(&page), [WHITE; 8]);
+}

crates/render/tests/it/main.rs

@@ -3,6 +3,7 @@
 mod cid_cff_unusable_font_dict;
 mod cid_fallback_query;
 mod fill_stroke_knockout;
+mod indexed_lookup;
 mod inline_image_ei;
 mod jbig2_invariants;
 mod knockout_backdrop;

vendor/README.md

@@ -39,7 +39,7 @@ Other resource categories retain their existing parent-chain lookup.
 
 | Crate | Version | Licence | Patches | Test |
 |---|---|---|---|---|
-| hayro-interpret | 0.7.0 | Apache-2.0 OR MIT | (1) select `/AP /N` appearance states via `/AS` (checkboxes, radios, stateful icons) and honour the NoView flag; (2) Type3 Unicode fallback to `/Encoding` glyph names, then printable ASCII; (3) `Type3Glyph::advance_width`; (4) `InterpreterSettings::ocg_overrides` for viewer layer toggles; (5) soft masks whose group lacks `/CS` were dropped (Chrome gradient text); (6) paint nesting cap `MAX_PAINT_NESTING` for tiling patterns and Type 3 glyphs (self-referencing content overflowed the stack); (7) CID width ranges in `/W` and `/W2` clamped to `MAX_CID` (a `0 4294967295 w` range hung); (8) `InterpreterSettings::hide_comments` skips markup annotations but keeps widgets and links (View ▸ Hide all comments); (9) `image_size_ok`: images and masks over 2^28 pixels are not decoded (a fuzzed `/W 4294967295` stencil mask allocated 4 GiB); (10) `MAX_NESTED_PAINTS`: nested paints (forms, Type 3 glyphs and tiling patterns inside one another) share a per-page budget of 50,000 (a glyph showing eight glyphs of its own font, sixteen deep, hung); (11) depends on skrifa 0.47 instead of 0.42 (upstream read-fonts fixed a Type 1 tokenizer that recursed through runs of numbers: a fuzzed font hung or overflowed the stack), with the one API change (`Charset::glyph_id` returns `Option`); (12) annotation appearances whose `/BBox` or `/Matrix` collapses to a line or a point are skipped (mapping them onto `/Rect` divided by zero and handed the device infinite or NaN matrices); (13) a Highlight annotation without a usable normal appearance is drawn from its `/QuadPoints`, `/C` and `/CA` with Multiply blending, as pdfcraft-annot draws highlights (it was not drawn at all); (14) simple-font `/Widths` (`read_widths`): entries only for codes 0–255 (one entry per code below `/FirstChar` was allocated: `4294967295` asked for 34 GB), an entry that isn't a number leaves its code missing instead of ending the array, and a `/LastChar` below `/FirstChar` gives no widths instead of discarding the font for a standard one; (15) the appearance box is mapped onto `/Rect` corner to corner: the translation now accounts for the scale (an appearance whose box didn't start at the origin and differed in size from `/Rect` was drawn shifted); (16) a clip set by `W`/`W*` takes effect after whichever path-painting operator ends the path (`S`, `f`, `B`, … as well as `n`; ISO 32000-2 §8.5.4): it was applied only on `n`, so `re W* S` clipped nothing and the clip stayed pending for a later, unrelated `n`; (17) `InterpreterSettings::cancelled`: once set, interpretation stops before the next content operator (a render pool sets it for every worker when dropped, after an edit or on closing, and for one worker when the watchdog gives up on its render, so renders whose answers nobody can receive stop instead of running to the end); (18) `MAX_PAGE_CONTENT`: the content a page holds at once (its contents, and the forms, tiling patterns and Type 3 glyph procedures being painted) stays within 512 MiB (`decode_content`; once it runs out, `InterpreterWarning::ContentTruncated` is reported (once per page by pdfcraft-render) so a truncated page is not a clean render; each level gives its bytes back when it finishes): forms nest fifty deep and each level kept its content while the next ran, so a form painting itself around a stream inflating to 256 MiB would have held over 12 GB; and Type 0 sampled functions take at most `MAX_SAMPLE_ENTRIES` (2^21) table points and `MAX_SAMPLE_VALUES` (2^23) samples, read no further than `/Size` asks (256 MiB of 1-bit samples asked for over 8 GB); (19) mesh shadings (types 4–7) are sampled only inside the area being drawn (`ShadingPattern::encode_within`, which the vendored hayro renderer calls with that area plus a pixel), within 8 visits per pixel of it plus 4 per triangle (`TriangleSampler`), and patches are triangulated one at a time: points far off the page were sampled over their whole bounding box, up to 65536 × 65536 entries (a fuzzed 147-byte patch stream; a synthetic page passed 6 GB in six seconds), ten thousand overlapping triangles ran for minutes, and every patch became 722 triangles before any limit; meshes hold at most `MAX_MESH_PATCHES` (2^16) patches and `MAX_MESH_TRIANGLES` (2^20) triangles, a lattice needs two to 2^19 vertices per row (`/VerticesPerRow 0` pushed empty rows for ever; one huge row read the whole stream), and a page samples at most `MAX_PAGE_MESH_TRIANGLES` (2^25) mesh triangles in all (a mesh painted a thousand times was triangulated a thousand times); each cut logs a warning; (20) PDF function construction is limited to 64 levels and 10,000 total nodes with cycle detection; calculator parsing and evaluation share procedure depth/work budgets, integer division/bitshift/rotation/index are panic-free, and fixed-size stack errors propagate; (21) a Type 0 font drawn with a substitute that isn't CID-keyed (it isn't embedded, or its program doesn't load) draws `.nLine truncated
+| hayro-interpret | 0.7.0 | Apache-2.0 OR MIT | (1) select `/AP /N` appearance states via `/AS` (checkboxes, radios, stateful icons) and honour the NoView flag; (2) Type3 Unicode fallback to `/Encoding` glyph names, then printable ASCII; (3) `Type3Glyph::advance_width`; (4) `InterpreterSettings::ocg_overrides` for viewer layer toggles; (5) soft masks whose group lacks `/CS` were dropped (Chrome gradient text); (6) paint nesting cap `MAX_PAINT_NESTING` for tiling patterns and Type 3 glyphs (self-referencing content overflowed the stack); (7) CID width ranges in `/W` and `/W2` clamped to `MAX_CID` (a `0 4294967295 w` range hung); (8) `InterpreterSettings::hide_comments` skips markup annotations but keeps widgets and links (View ▸ Hide all comments); (9) `image_size_ok`: images and masks over 2^28 pixels are not decoded (a fuzzed `/W 4294967295` stencil mask allocated 4 GiB); (10) `MAX_NESTED_PAINTS`: nested paints (forms, Type 3 glyphs and tiling patterns inside one another) share a per-page budget of 50,000 (a glyph showing eight glyphs of its own font, sixteen deep, hung); (11) depends on skrifa 0.47 instead of 0.42 (upstream read-fonts fixed a Type 1 tokenizer that recursed through runs of numbers: a fuzzed font hung or overflowed the stack), with the one API change (`Charset::glyph_id` returns `Option`); (12) annotation appearances whose `/BBox` or `/Matrix` collapses to a line or a point are skipped (mapping them onto `/Rect` divided by zero and handed the device infinite or NaN matrices); (13) a Highlight annotation without a usable normal appearance is drawn from its `/QuadPoints`, `/C` and `/CA` with Multiply blending, as pdfcraft-annot draws highlights (it was not drawn at all); (14) simple-font `/Widths` (`read_widths`): entries only for codes 0–255 (one entry per code below `/FirstChar` was allocated: `4294967295` asked for 34 GB), an entry that isn't a number leaves its code missing instead of ending the array, and a `/LastChar` below `/FirstChar` gives no widths instead of discarding the font for a standard one; (15) the appearance box is mapped onto `/Rect` corner to corner: the translation now accounts for the scale (an appearance whose box didn't start at the origin and differed in size from `/Rect` was drawn shifted); (16) a clip set by `W`/`W*` takes effect after whichever path-painting operator ends the path (`S`, `f`, `B`, … as well as `n`; ISO 32000-2 §8.5.4): it was applied only on `n`, so `re W* S` clipped nothing and the clip stayed pending for a later, unrelated `n`; (17) `InterpreterSettings::cancelled`: once set, interpretation stops before the next content operator (a render pool sets it for every worker when dropped, after an edit or on closing, and for one worker when the watchdog gives up on its render, so renders whose answers nobody can receive stop instead of running to the end); (18) `MAX_PAGE_CONTENT`: the content a page holds at once (its contents, and the forms, tiling patterns and Type 3 glyph procedures being painted) stays within 512 MiB (`decode_content`; once it runs out, `InterpreterWarning::ContentTruncated` is reported (once per page by pdfcraft-render) so a truncated page is not a clean render; each level gives its bytes back when it finishes): forms nest fifty deep and each level kept its content while the next ran, so a form painting itself around a stream inflating to 256 MiB would have held over 12 GB; and Type 0 sampled functions take at most `MAX_SAMPLE_ENTRIES` (2^21) table points and `MAX_SAMPLE_VALUES` (2^23) samples, read no further than `/Size` asks (256 MiB of 1-bit samples asked for over 8 GB); (19) mesh shadings (types 4–7) are sampled only inside the area being drawn (`ShadingPattern::encode_within`, which the vendored hayro renderer calls with that area plus a pixel), within 8 visits per pixel of it plus 4 per triangle (`TriangleSampler`), and patches are triangulated one at a time: points far off the page were sampled over their whole bounding box, up to 65536 × 65536 entries (a fuzzed 147-byte patch stream; a synthetic page passed 6 GB in six seconds), ten thousand overlapping triangles ran for minutes, and every patch became 722 triangles before any limit; meshes hold at most `MAX_MESH_PATCHES` (2^16) patches and `MAX_MESH_TRIANGLES` (2^20) triangles, a lattice needs two to 2^19 vertices per row (`/VerticesPerRow 0` pushed empty rows for ever; one huge row read the whole stream), and a page samples at most `MAX_PAGE_MESH_TRIANGLES` (2^25) mesh triangles in all (a mesh painted a thousand times was triangulated a thousand times); each cut logs a warning; (20) PDF function construction is limited to 64 levels and 10,000 total nodes with cycle detection; calculator parsing and evaluation share procedure depth/work budgets, integer division/bitshift/rotation/index are panic-free, and fixed-size stack errors propagate; (21) a Type 0 font drawn with a substitute that isn't CID-keyed (it isn't embedded, or its program doesn't load) draws `.nLine truncated
 | hayro | 0.7.1 | Apache-2.0 OR MIT | (1) `RenderSettings::{x_offset, y_offset}` to render one tile of a large page; (2) images over `MAX_IMAGE_PIXELS` (2^28) are skipped (a fuzzed `/W 4294967295` hung in resampling); (3) `tiling_cell_scale` keeps a tiling pattern's cell pixmap within 3000 px a side whatever its `/XStep`/`/YStep` (a fuzzed step far beyond `/BBox` allocated 2 GB); (4) stroke widths clamped to 4 × (canvas width + height) device pixels (a fuzzed `/LW 9223372036854775807` allocated 10 GB in stroke expansion); (5) `usable_dash_pattern`: a stroke whose dash pattern could cut it into more than `MAX_DASHES_PER_STROKE` (250,000) pieces, counting kurbo's restart of the pattern at every subpath and `ClosePath`, is drawn solid (a fuzzed `/D [[11] 0]` on a line from x = 92234775807 allocated over 5 GB, `[0.000001] 0 d` on a 20 pt line over 10 GB); (6) a dash pattern with a negative or non-finite entry, or a period that isn't positive, is drawn solid (`[-1 -1] 0 d` looped for ever in kurbo); (7) `render_size` and `render_into`: render into a buffer the caller allocates (PdfCraft renders into pixel-aligned memory that the GUI takes over as texture data); `render` is `render_into` on a new pixmap; (8) checked image source/target/RGBA, alpha-premultiplication scratch and intermediate sizes before resampling, finite transforms and exact source/mask buffers; sources wider than u16 may shrink before backend limits, with a 2^20 source-axis cap and conservative 128 MiB filter-planning estimate; an over-limit or failed resize keeps the original image resolution and geometry when the source fits the backend, otherwise skips it, with a logged warning; staging, resampling output/scratch, RGBA conversion and padding buffers use fallible reserves; (9) `may_paint_canvas`: a fill or stroke whose control-point box, grown by how far its paint reaches past the path (half the stroke width times the larger of the miter limit and √2) plus 0.5 user units and 2 pixels, misses the canvas is not drawn, so a tile no longer strokes and fills every path of its page (item 5 of #623); `set_skip_offscreen_paths` turns this off for the regression test; (10) `/Interpolate` governs the pre-resize too (#624): a minified image that does not request interpolation is pre-resized with nearest-neighbour (since (21), only inside Type 3 glyphs), planned without the Catmull-Rom filter limits (`image_resampling_size` takes the flag); a Type 3 glyph image is sampled bicubically only when interpolated; a minified image with a matching soft mask is nearest-sampled straight into premultiplied RGBA; (11) `may_paint_rect`: an image whose stream-size box, mapped to the device, misses the canvas is skipped before decoding (decoded data can only shrink against that box), so a tile of an image-heavy page no longer decodes, resamples and rasterizes every image on it; (12) preview renders (`InterpreterSettings::preview`: backdrops, thumbnails, grid cells, signature-drag backgrounds) skip images under `PREVIEW_MIN_IMAGE_AREA` device pixels, while final renders always draw them; (13) reading colours (#897): `RenderSettings::color_replacement` takes a `ColorReplacement` (text colour, background, `only_black`, `line_art`) that remaps solid text and line-art paint while rendering, glyph outlines counting as text and stencil-mask ink as line art; images, image alpha masks, soft masks and pattern cells keep their colours; `render_pdf` gains a `color_replacement: Option<ColorReplacement>` argument (pass `None` for true colours); the upstream-only `[dev-dependencies.log]` entry is removed from `Cargo.toml` (log is a normal dependency since the logging patch), so `cargo test -p hayro` runs this crate's unit tests; (14) a shrunk image that is interpolated (or, since (21), area-averaged) and whose axes lie along the device's (upright, mirrored or a quarter turn, as on a page under `/Rotate`) is snapped to whole device pixels: shrunk once to its rounded device rectangle (halves round up, so a tile snaps an edge where the whole page does) and drawn 1:1 from the rounded corner, instead of a ceil-sized raster resampled again at a fractional scale and offset (`snap_minified_image`; scans looked soft next to Acrobat, and until #5530 a turned one still did, up to 200 grey levels off the same page drawn upright). An image counts as axis-aligned when the terms the snap drops move its far edges by at most a hundredth of a device pixel (`SNAP_SKEW`), not by a tolerance on the coefficients, which let a slight skew of a very wide image through; and the snapped 1:1 transform is used only when the shrink produced the snapped size, so a failed resize keeps the full transform instead of drawing the source raster at its source size; (15) a soft mask of another size than its image is filtered when shrunk, whatever its `/Interpolate` (#1649): such a mask carries the shape on its own (Word saves text as a tiny black image with a glyph-bitmap mask), and nearest-neighbour dropped every stroLine truncated
 | hayro-jbig2 | 0.3.0 | Apache-2.0 OR MIT | (1) at most 2^28 pixels per bitmap and page bitmap, and 2^25 cells per halftone grid and grayscale image (each side could be 65535: 4.3 billion pixels decoded one by one, and a 4-byte value per cell); (2) at most 2^26 symbol pixels per symbol dictionary (a fuzzed dictionary decoded for minutes); (3) a symbol dictionary's height classes are capped at one per new symbol and its export runs at `2 × symbols + 1`, each plus `EXTRA_EMPTY_STEPS` (1024): an empty height class or a zero-length export run doesn't advance, and past the end of its data the arithmetic decoder can decode them for ever (one changed byte hung the renderer); (4) a segment header's long-form referred-to segment count is capped at `u16::MAX` and a larger count is refused before its retention or referred-to bytes are read: the 29-bit count went straight to `Vec::with_capacity`, a reservation of about 2 GiB whose failure aborts the process (from storytold/craft-libs#176); (5) what a page's segments hold for later ones (intermediate regions, pattern and symbol dictionaries with their headers and retained contexts, and Huffman tables) is at most 64 MiB, two square bitmaps of 16,384 a side (`MAX_RETAINED_WORDS`), and a dictionary's new and exported symbols are counted against it as they are made: each was bounded on its own, but a few dozen bytes an intermediate region held 33 MB each, and 8 MiB symbols exported again and again killed the process, so a few kilobytes could exhaust memory; (6) a segment may refer to at most 2^20 symbols, checked before any are collected: references may repeat, and 65,535 of them to one dictionary, in a 78 KB stream, asked for 17 GB; (7) a pattern dictionary keeps at most 16 MiB of shifted copies, which only speed up drawing, and none that wouldn't fit what the page has left (`MAX_SHIFTED_WORDS`); (8) a custom Huffman table has at most 2^16 lines, and a stream with its globals at most 2^20 segments and 2^22 references, checked as they are read: both grew before anything counted them, and PDFs of a kilobyte or so reached 5 GB and more behind nested Flate; (9) a page past the bounds of (1) is refused when the stream is opened: a caller sizes its output from the header before the page bitmap is made, and an Indexed image in a 743-byte PDF asked for 137 GB, which aborts; (10) a stream may ask for at most 2^30 units of decoding work, four of the largest pages, about 6 s to 17 s (`MAX_WORK`): decoding a bitmap costs a pixel each and 4 a row, copying one a word each and 1 a row, and each integer an item reads costs 38, what one takes at worst, so one with no pixels pays too, before it is decoded: each bitmap was bounded, but nothing bounded how many there were, and a 795-byte PDF took 265 s; (11) `Image::decode_within` spends from what a run of images may still do, starting with its page's pixels, and `charge_reading` charges reading a stream and its globals before anything is parsed, so an operation over a document can bound what `MAX_WORK` bounds for one image, and images that fail or decode nothing add up too: a file of thousands of images made to reach that limit took as many times as long; (12) adding to a location, a reference offset, a symbol's corner or a Huffman range line near the end of an `i32`, and negating a decoded integer of 2^31, overflowed: a panic wherever overflow is checked. The refinement context, a text region's corners, the integer decoder and the Huffman table wrap, as a release build did, and `Bitmap::combine` saturates, which comes to the same, as a location that overflows is past the end of any page. What is drawn doesn't change; (13) a bitmap filled with a default pixel of 1 (a page, or an intermediate text or halftone region) had 1s past its width in the last word of each row, and a refinement, which reads its reference a word at a time, read them as pixels where the standard has 0 (6.3.5.2): refining such a reference not a multiple of 32 wide decoded to other pixels than were coded, release builds too. `Bitmap::clear_padding` clears those bits when a bitmap is filled with 1s, as cutting symbols and patterns from a collective bitmap already did. Ghostscript reads the test's page and text streams as they decode now; (14) a region's location is unsigned (5.4.4), so one past `i32::MAX` is off the page, where the decoder took it for a signed one: a region there was drawn back onto the page from the left or the top (at 0xFFFFFFFC, four pixels left of it), and a refinement there, of the page or of a region at the same place (7.4.7.5), was refused as an overflow, and the page with it. `Bitmap::combine_region` keeps such a location past the page. A refinement of another region refines it where it is, as Table 35 has it (GRREFERENCEDX and GRREFERENCEDY 0), whatever the two locations say; a refinement of the page, see (16); (15) a refinement whose reference is offset right of it (GRREFERENCEDX above 0, which a refined text instance has from a wider bitmap, TaLine truncated
 | lopdf | 0.45.0 | MIT | (1) PNG predictor rows longer than the data are refused before their buffers are allocated (a fuzzed `/Columns 4294967295` allocated two 4 GiB rows), and empty frames return without row buffers; (2) predictor row widths use saturating arithmetic or reject overflow; (3) sub-byte TIFF Predictor 2 scratch is bounded by the samples present in the input instead of the declared `/Colors`, aligned with [lopdf#589](https://github.com/J-F-Liu/lopdf/pull/589); (4) the object lookup keeps its cycle history through compressed objects, requires an ordinary in-file container, and caps the chain of indirect lookups at 16 (a chain of new objects, each a `/Length` compressed in a container whose own `/Length` names the next, or a stream whose `/Length` names the next stream, overflowed the stack: a dev build's 2 MiB thread within 800 hops); (5) object-stream members are parsed within their own extent (up to the next distinct offset), an entry whose offset repeats an earlier entry's is skipped (so a repeated object number can resolve to its earlier entry), and offsets are processed in index order; (6) `find_latest_trailer`: a rebuild takes a `/Type /XRef` stream dictionary as the trailer when the file has no `trailer` keyword (the latest in the file wins, as a later `trailer` would), so a concatenated or truncated file keeps its /Info and /ID instead of failing to load; a candidate whose /Root is the catalog beats one whose /Root is another dictionary, and an in-file /Root must match its generation (#2365) | (1) `inspect::tests::png_predictor_rows_longer_than_the_data_are_refused_before_allocating`, `inspect::tests::empty_png_predictor_frames_do_not_allocate_rows`; (2) `inspect::tests::tiff_subbyte_predictor_row_width_overflow_is_refused`; (3) `inspect::tests::tiff_subbyte_predictor_scratch_is_bounded_by_available_samples`, `inspect::tests::tiff_subbyte_predictors_keep_components_rows_and_padding`; (4) `inspect::tests::compressed_object_self_container_does_not_recurse`, `inspect::tests::compressed_object_mutual_containers_do_not_recurse`, `inspect::tests::object_stream_length_cycle_does_not_recurse`, `inspect::tests::compressed_object_container_chain_does_not_recurse`, `inspect::tests::compressed_length_chain_does_not_recurse`, `inspect::tests::plain_stream_length_chain_does_not_recurse`, `inspect::tests::compressed_objects_and_indirect_stream_length_still_load`, `inspect::tests::compressed_length_in_a_container_with_an_indirect_length_still_loads`; (5) `inspect::tests::object_stream_member_extents_bound_owned_strings`, `inspect::tests::object_stream_member_extents_preserve_index_order_and_whitespace`; (6) `inspect::tests::rebuilt_file_keeps_the_info_of_its_cross_reference_stream`, `inspect::tests::rebuilt_file_prefers_a_trailer_whose_root_is_the_catalog` |

vendor/hayro-interpret/src/cache.rs

@@ -10,9 +10,10 @@ use std::sync::{Arc, Mutex};
 
 type CacheMap = HashMap<u128, Option<Box<dyn Any + Send + Sync>>>;
 /// The objects read from a document, by key. PdfCraft patch (48): the second field counts the
-/// ICC profiles read through it ([`Cache::icc_profile_reads`]).
+/// ICC profiles read through it ([`Cache::icc_profile_reads`]). PdfCraft patch (80): the third
+/// counts the `/Indexed` lookup tables read ([`Cache::indexed_lookup_reads`]).
 #[derive(Clone)]
-pub(crate) struct Cache(Arc<Mutex<CacheMap>>, Arc<AtomicUsize>);
+pub(crate) struct Cache(Arc<Mutex<CacheMap>>, Arc<AtomicUsize>, Arc<AtomicUsize>);
 impl Default for Cache {
     fn default() -> Self {
         Self::new()
@@ -21,7 +22,11 @@ impl Default for Cache {
 
 impl Cache {
     pub(crate) fn new() -> Self {
-        Self(Arc::new(Mutex::new(HashMap::new())), Arc::default())
+        Self(
+            Arc::new(Mutex::new(HashMap::new())),
+            Arc::default(),
+            Arc::default(),
+        )
     }
 
     /// PdfCraft patch (48): count one ICC profile read, the colour space of an `/ICCBased`
@@ -35,6 +40,17 @@ impl Cache {
         self.1.load(Ordering::Relaxed)
     }
 
+    /// PdfCraft patch (80): count one `/Indexed` lookup table read (a stream decoded or a
+    /// string copied).
+    pub(crate) fn count_indexed_lookup_read(&self) {
+        self.2.fetch_add(1, Ordering::Relaxed);
+    }
+
+    /// PdfCraft patch (80): how many `/Indexed` lookup tables were read through this cache.
+    pub(crate) fn indexed_lookup_reads(&self) -> usize {
+        self.2.load(Ordering::Relaxed)
+    }
+
     pub(crate) fn get_or_insert_with<T: Clone + Send + Sync + 'static>(
         &self,
         id: u128,
@@ -142,6 +158,7 @@ fn warning_bit(warning: InterpreterWarning) -> u8 {
         InterpreterWarning::ContentTruncated => 4,
         InterpreterWarning::ColorFallback => 8,
         InterpreterWarning::Type3WidthsMissing => 16,
+        InterpreterWarning::MalformedIndexedImage => 32,
     }
 }
 
@@ -153,6 +170,7 @@ fn warnings_in(bits: u8) -> impl Iterator<Item = InterpreterWarning> {
         InterpreterWarning::ContentTruncated,
         InterpreterWarning::ColorFallback,
         InterpreterWarning::Type3WidthsMissing,
+        InterpreterWarning::MalformedIndexedImage,
     ]
     .into_iter()
     .filter(move |warning| bits & warning_bit(*warning) != 0)

vendor/hayro-interpret/src/color.rs

@@ -258,7 +258,7 @@ impl ColorSpaceType {
                     return Some(Self::Lab(Lab::new(&lab_dict)?));
                 }
                 INDEXED | I => {
-                    return Some(Self::Indexed(Indexed::new(&color_array, cache, depth)?));
+                    return Some(Self::Indexed(Indexed::new(&color_array, cache, depth, false)?));
                 }
                 SEPARATION => {
                     return Some(Self::Separation(Separation::new(&color_array, cache, depth)?));
@@ -312,6 +312,22 @@ impl ColorSpace {
         Some(Self(Arc::new(ColorSpaceType::new(object, cache, depth)?)))
     }
 
+    /// PdfCraft patch (80): as `new`, except that an `/Indexed` space with no lookup table is
+    /// built (marked tableless) rather than unreadable, for the image and colour operators to
+    /// handle it (see `Indexed::new`). Every other space is read as by `new`.
+    pub(crate) fn new_allowing_tableless(object: Object<'_>, cache: &Cache) -> Option<Self> {
+        if let Object::Array(array) = &object
+            && array
+                .flex_iter()
+                .next::<Name<'_>>()
+                .is_some_and(|family| matches!(family.deref(), INDEXED | I))
+        {
+            return Indexed::new(array, cache, 0, true)
+                .map(|indexed| Self(Arc::new(ColorSpaceType::Indexed(indexed))));
+        }
+        Self::new(object, cache)
+    }
+
     /// Create a new color space from the name.
     pub(crate) fn new_from_name(name: &Name<'_>) -> Option<Self> {
         ColorSpaceType::new_from_name(name).map(|c| Self(Arc::new(c)))
@@ -354,6 +370,18 @@ impl ColorSpace {
         matches!(self.0.as_ref(), ColorSpaceType::Indexed(_))
     }
 
+    /// PdfCraft patch (80): whether this is an indexed colour space that broke ISO 32000-2
+    /// §8.6.6.3 and was recovered (see `Indexed::new`).
+    pub(crate) fn is_recovered_indexed(&self) -> bool {
+        matches!(self.0.as_ref(), ColorSpaceType::Indexed(i) if i.recovered && !i.tableless)
+    }
+
+    /// PdfCraft patch (80): whether this is an indexed colour space with no lookup table (see
+    /// `Indexed::new`), which only `ColorSpace::new_allowing_tableless` builds.
+    pub(crate) fn is_tableless_indexed(&self) -> bool {
+        matches!(self.0.as_ref(), ColorSpaceType::Indexed(i) if i.tableless)
+    }
+
     /// Get the default decode array for the color space.
     pub(crate) fn default_decode_arr(&self, n: f32) -> SmallVec<[(f32, f32); 4]> {
         match self.0.as_ref() {
@@ -1003,55 +1031,144 @@ impl ToRgb for Lab {
 
 #[derive(Debug, Clone)]
 pub(crate) struct Indexed {
+    /// `hival + 1` entries of the base's component count each; entries from `present` on are
+    /// zeros and are drawn black (see `convert_f32`).
     values: Vec<Vec<f32>>,
     hival: u8,
     base: Box<ColorSpace>,
+    /// PdfCraft patch (80): how many entries the lookup table holds in full (at most
+    /// `hival + 1`). An index at or past it is drawn black.
+    present: usize,
+    /// PdfCraft patch (80): the space broke ISO 32000-2 §8.6.6.3 and was recovered (see
+    /// `Indexed::new`).
+    recovered: bool,
+    /// PdfCraft patch (80): the space has no lookup table at all (see `Indexed::new`).
+    tableless: bool,
 }
 
 impl Indexed {
-    fn new(array: &Array<'_>, cache: &Cache, depth: u32) -> Option<Self> {
+    /// PdfCraft patch (80): ISO 32000-2 §8.6.6.3 says `hival` "shall be an integer" no greater
+    /// than 255 and the lookup table "shall be m × (hival + 1) bytes long", but not how a reader
+    /// recovers when a file breaks this. Most readers draw such a space, and so does this:
+    /// - a real `hival` is truncated toward zero (upstream read it so too);
+    /// - a `hival` past 255 is taken as 255;
+    /// - a lookup table shorter than m × (hival + 1) bytes keeps the entries it holds in full,
+    ///   and an index past them is drawn black, after conversion, whatever the base space (zero
+    ///   bytes would be white in DeviceCMYK, Separation and DeviceN, and blue in Lab).
+    ///
+    /// Any of these sets `recovered`, which the drawing paths report. A `hival` below zero (once
+    /// truncated), or not a number, has no reading: the space cannot be read, as upstream, and
+    /// `has_unusable_hival` lets the image path report it.
+    ///
+    /// A lookup that is missing, or neither a string nor a stream that decodes, is no table at
+    /// all. Upstream could not read such a space either. With `allow_tableless` it is built,
+    /// marked `tableless`, so that the image path can skip the image with a warning and the
+    /// colour operators can fall back as upstream; without it, it is `None` as upstream.
+    fn new(array: &Array<'_>, cache: &Cache, depth: u32, allow_tableless: bool) -> Option<Self> {
         let mut iter = array.flex_iter();
         // Skip name
         let _ = iter.next::<Name<'_>>()?;
         let base_color_space = ColorSpace::within(iter.next::<Object<'_>>()?, cache, depth + 1)?;
-        let hival = iter.next::<u32>()?.min(u8::MAX as u32) as u8;
+        let (hival, hival_recovered) = indexed_hival(iter.next::<object::Number>()?)?;
 
-        let values = {
-            // PdfCraft patch (71): read once, as whichever it is. A failed read as a stream leaves
-            // a direct string in place but uses up a reference, so a table in an indirect string
-            // was lost, and the space with it.
-            let data = match iter.next::<Object<'_>>()? {
-                Object::Stream(s) => s.decoded().ok()?,
-                Object::String(s) => Cow::Owned(s.to_vec()),
-                _ => return None,
-            };
+        // At most 255 components (a `u8`), and at most 256 entries.
+        let num_components = usize::from(base_color_space.num_components());
+        let entries = usize::from(hival) + 1;
 
-            let num_components = base_color_space.num_components();
-
-            let mut byte_iter = data.iter().copied();
-
-            let mut vals = vec![];
-            for _ in 0..=hival {
-                let mut temp = vec![];
-
-                for _ in 0..num_components {
-                    temp.push(byte_iter.next()? as f32 / 255.0);
-                }
-
-                vals.push(temp);
-            }
-
-            vals
-        };
+        // PdfCraft patch (80): read once and then by type, so a string lookup given by an
+        // indirect reference is found (a failed stream read had already consumed the
+        // reference).
+        let data = iter
+            .next::<Object<'_>>()
+            .and_then(|o| indexed_lookup(o, cache));
+        let tableless = data.is_none();
+        if tableless && !allow_tableless {
+            return None;
+        }
+        let data: &[u8] = data.as_deref().unwrap_or_default();
+        let present = data
+            .len()
+            .checked_div(num_components)
+            .unwrap_or(0)
+            .min(entries);
+        let mut bytes = data.iter().copied();
+        let values: Vec<Vec<f32>> = (0..entries)
+            .map(|entry| {
+                (0..num_components)
+                    .map(|_| {
+                        if entry < present {
+                            f32::from(bytes.next().unwrap_or(0)) / 255.0
+                        } else {
+                            0.0
+                        }
+                    })
+                    .collect()
+            })
+            .collect();
 
         Some(Self {
             values,
             hival,
             base: Box::new(base_color_space),
+            present,
+            recovered: hival_recovered || present < entries,
+            tableless,
         })
     }
 }
 
+/// PdfCraft patch (80): an Indexed space's `hival` (ISO 32000-2 §8.6.6.3) as drawn, and whether
+/// it had to be recovered: a real is truncated toward zero and a value past 255 is taken as 255.
+/// `None` for a value below zero once truncated, or one that is not finite.
+fn indexed_hival(number: object::Number) -> Option<(u8, bool)> {
+    let value = number.as_f64();
+    let whole = value.trunc();
+    // NaN fails this test too.
+    if !(whole >= 0.0 && whole.is_finite()) {
+        return None;
+    }
+    let clamped = whole.min(f64::from(u8::MAX));
+    // `clamped` is a whole number in 0..=255, so the cast is exact.
+    Some((clamped as u8, clamped != value))
+}
+
+/// PdfCraft patch (80): whether `object` is an `/Indexed` colour space array whose `hival` has no
+/// reading (below zero, or not a number), which `Indexed::new` cannot recover. Reads the array's
+/// first and third entries only; the lookup table is not touched.
+pub(crate) fn has_unusable_hival(object: &Object<'_>) -> bool {
+    let Object::Array(array) = object else {
+        return false;
+    };
+    let mut iter = array.flex_iter();
+    if !iter
+        .next::<Name<'_>>()
+        .is_some_and(|family| matches!(family.deref(), INDEXED | I))
+    {
+        return false;
+    }
+    let _base = iter.next::<Object<'_>>();
+    iter.next::<object::Number>()
+        .and_then(indexed_hival)
+        .is_none()
+}
+
+/// PdfCraft patch (80): an Indexed space's lookup table, a byte string or a stream (ISO 32000-2
+/// §8.6.6.3), with a stream's filters applied. `None` for any other object or a stream that
+/// cannot be decoded. A stream that decodes to nothing although it holds data counts as one that
+/// cannot be decoded: the Flate decoder recovers what it can from a broken stream, and here
+/// recovered nothing. Each call is counted on `cache` (`Cache::indexed_lookup_reads`).
+fn indexed_lookup<'a>(lookup: Object<'a>, cache: &Cache) -> Option<Cow<'a, [u8]>> {
+    cache.count_indexed_lookup_read();
+    match lookup {
+        Object::Stream(s) => s
+            .decoded()
+            .ok()
+            .filter(|d| !d.is_empty() || s.raw_data().is_empty()),
+        Object::String(s) => Some(Cow::Owned(s.to_vec())),
+        _ => None,
+    }
+}
+
 impl ToRgb for Indexed {
     fn convert_f32(&self, input: &[f32], output: &mut [u8], _: bool) -> Option<()> {
         let mut indexed = vec![0.0; input.len() * self.base.num_components() as usize];
@@ -1065,7 +1182,20 @@ impl ToRgb for Indexed {
             output.copy_from_slice(&self.values[idx]);
         }
 
-        self.base.convert_f32(&indexed, output, true)
+        self.base.convert_f32(&indexed, output, true)?;
+
+        // PdfCraft patch (80): an index past the entries the table holds is drawn black, in
+        // whatever space the base is (see `Indexed::new`).
+        if self.present <= usize::from(self.hival) {
+            for (input, rgb) in input.iter().copied().zip(output.chunks_exact_mut(3)) {
+                let idx = (input.clamp(0.0, self.hival as f32) + 0.5) as usize;
+                if idx >= self.present {
+                    rgb.fill(0);
+                }
+            }
+        }
+
+        Some(())
     }
 }
 

vendor/hayro-interpret/src/context.rs

@@ -220,6 +220,13 @@ impl<'a> InterpreterCache<'a> {
     pub fn icc_profile_reads(&self) -> usize {
         self.object_cache.icc_profile_reads()
     }
+
+    /// PdfCraft patch (80): how many `/Indexed` lookup tables were read through this cache, for
+    /// diagnostics and tests. A colour space named by a resource is built, and its table read,
+    /// once however many images or colours use it.
+    pub fn indexed_lookup_reads(&self) -> usize {
+        self.object_cache.indexed_lookup_reads()
+    }
 }
 
 /// A per-page interpretation context that borrows shared data from an [`InterpreterCache`].
@@ -565,6 +572,8 @@ impl<'a> Context<'a> {
         self.get_mut().ctm *= transform;
     }
 
+    /// PdfCraft patch (80): an `/Indexed` resource with no lookup table is built, marked
+    /// tableless, for the image and colour operators to handle (see `Indexed::new`).
     pub(crate) fn get_color_space(
         &mut self,
         resources: &Resources<'_>,
@@ -574,7 +583,10 @@ impl<'a> Context<'a> {
         self.interpreter_cache
             .object_cache
             .get_or_insert_with(cs_object.cache_key(), || {
-                ColorSpace::new(cs_object.clone(), &self.interpreter_cache.object_cache)
+                ColorSpace::new_allowing_tableless(
+                    cs_object.clone(),
+                    &self.interpreter_cache.object_cache,
+                )
             })
     }
 

vendor/hayro-interpret/src/interpret/mod.rs

@@ -166,6 +166,11 @@ pub enum InterpreterWarning {
     /// read, or a Separation or DeviceN tint transform that cannot be evaluated) and was painted
     /// in a fallback colour. Reported at most once per content stream.
     ColorFallback,
+    /// PdfCraft patch (80): an image's `/Indexed` colour space is malformed (ISO 32000-2
+    /// §8.6.6.3). The image was drawn recovered (indices past a short lookup table in black, a
+    /// `hival` past 255 taken as 255, a real one truncated), or, with a `hival` below zero or not a
+    /// number, in DeviceGray; with no lookup table at all it was not drawn.
+    MalformedIndexedImage,
     /// PdfCraft patch (90): a Type 3 font without `/Widths` (or `/FirstChar`/`/LastChar`),
     /// which ISO 32000-2 Table 110 requires, was selected; its glyphs advance by the widths
     /// their glyph procedures set with `d0`/`d1`. Reported each time such a font is selected.
@@ -695,11 +700,19 @@ fn interpret_window<'a>(
                 // PdfCraft patch (38): see `named_color_space`.
                 // PdfCraft patch (55): a space that cannot be read is drawn in DeviceGray, as
                 // before, and the fallback is reported.
+                // PdfCraft patch (80): an Indexed space with no lookup table counts as one that
+                // cannot be read, as upstream.
                 let cs = named_color_space(c.0, resources, &defaults, context)
+                    .filter(|cs| !cs.is_tableless_indexed())
                     .unwrap_or_else(|| {
                         crate::color::note_color_fallback();
                         ColorSpace::device_gray()
                     });
+                // PdfCraft patch (80): so is an Indexed space drawn recovered (indices past a
+                // short table drawn black, a `hival` clamped or truncated; see `Indexed::new`).
+                if cs.is_recovered_indexed() {
+                    crate::color::note_color_fallback();
+                }
 
                 if !cs.is_pattern() {
                     context.get_mut().graphics_state.stroke_pattern = None;
@@ -711,11 +724,19 @@ fn interpret_window<'a>(
                 // PdfCraft patch (38): see `named_color_space`.
                 // PdfCraft patch (55): a space that cannot be read is drawn in DeviceGray, as
                 // before, and the fallback is reported.
+                // PdfCraft patch (80): an Indexed space with no lookup table counts as one that
+                // cannot be read, as upstream.
                 let cs = named_color_space(c.0, resources, &defaults, context)
+                    .filter(|cs| !cs.is_tableless_indexed())
                     .unwrap_or_else(|| {
                         crate::color::note_color_fallback();
                         ColorSpace::device_gray()
                     });
+                // PdfCraft patch (80): so is an Indexed space drawn recovered (indices past a
+                // short table drawn black, a `hival` clamped or truncated; see `Indexed::new`).
+                if cs.is_recovered_indexed() {
+                    crate::color::note_color_fallback();
+                }
 
                 if !cs.is_pattern() {
                     context.get_mut().graphics_state.non_stroke_pattern = None;
@@ -971,6 +992,24 @@ fn interpret_window<'a>(
                 let warning_sink = context.settings.warning_sink.clone();
                 let transfer_function = context.get().graphics_state.transfer_function.clone();
                 let cache = context.interpreter_cache.object_cache.clone();
+                // PdfCraft patch (80): an inline image naming a resource colour space that is an
+                // Indexed space whose `hival` has no reading is reported, as `ImageXObject::new`
+                // reports one given in place; the image falls back as upstream. Only the array's
+                // name and `hival` are read here, never its lookup table: a recovered space is
+                // reported by `ImageXObject::new` from the space it builds through the resource
+                // cache. A device space name (or its inline abbreviation) never refers to a
+                // resource (ISO 32000-2 §8.6.8), and `ImageXObject::new` takes it first, so a
+                // resource of that name is not consulted.
+                let inline_dict = i.0.dict();
+                if inline_dict
+                    .get::<hayro_syntax::object::Name<'_>>(hayro_syntax::object::dict::keys::CS)
+                    .or_else(|| inline_dict.get::<hayro_syntax::object::Name<'_>>(hayro_syntax::object::dict::keys::COLORSPACE))
+                    .filter(|name| ColorSpace::new_from_name(name).is_none())
+                    .and_then(|name| resources.get_color_space(&name))
+                    .is_some_and(|cs| crate::color::has_unusable_hival(&cs))
+                {
+                    warning_sink(InterpreterWarning::MalformedIndexedImage);
+                }
                 // PdfCraft patch: inline images are not cached (see `DecodedImageKey`).
                 // PdfCraft patch (38): an inline image in a device space, named directly or by a
                 // resource, draws in its default.

vendor/hayro-interpret/src/x_object.rs

@@ -526,6 +526,7 @@ impl<'a> ImageXObject<'a> {
         is_mask |= is_stencil_mask;
 
         let mut default_space = None;
+        let mut malformed_indexed = false;
         let image_cs = if is_mask {
             // Masks are always single-channel.
             Some(ColorSpace::device_gray())
@@ -543,9 +544,14 @@ impl<'a> ImageXObject<'a> {
                 default_space = Some(key);
                 Some(space)
             } else {
+                // PdfCraft patch (80): an Indexed space given in place whose `hival` has no
+                // reading cannot be built and falls back below, as upstream; it is reported.
+                malformed_indexed = cs_obj
+                    .as_ref()
+                    .is_some_and(crate::color::has_unusable_hival);
                 cs_obj
                     .clone()
-                    .and_then(|c| ColorSpace::new(c, cache))
+                    .and_then(|c| ColorSpace::new_allowing_tableless(c, cache))
                     // Inline images can also refer to color spaces by name.
                     .or_else(|| {
                         cs_obj
@@ -554,6 +560,24 @@ impl<'a> ImageXObject<'a> {
                     })
             }
         };
+        // PdfCraft patch (80): an Indexed space that ISO 32000-2 §8.6.6.3 rules out (a short
+        // lookup table, a `hival` past 255 or with a fraction) is drawn recovered (see
+        // `Indexed::new`), and reported. The check asks the space that was built, so it reads
+        // no lookup table of its own (an inline image naming a resource builds it once, through
+        // the resource cache).
+        malformed_indexed |= image_cs
+            .as_ref()
+            .is_some_and(ColorSpace::is_recovered_indexed);
+        // PdfCraft patch (80): an Indexed space with no lookup table at all (missing, or not a
+        // string or a stream that decodes) leaves nothing to draw: the image is skipped, and
+        // reported. Upstream drew it in DeviceGray.
+        if image_cs
+            .as_ref()
+            .is_some_and(ColorSpace::is_tableless_indexed)
+        {
+            warning_sink(InterpreterWarning::MalformedIndexedImage);
+            return None;
+        }
 
         let interpolate = dict
             .get::<bool>(I)
@@ -567,6 +591,10 @@ impl<'a> ImageXObject<'a> {
             return None;
         }
 
+        if malformed_indexed {
+            warning_sink(InterpreterWarning::MalformedIndexedImage);
+        }
+
         Some(Self {
             width,
             cache: cache.clone(),