Currently the `Catalog.metadata` getter only handles errors during parsing, however in a *corrupt* PDF document fetching of the raw /Metadata can obviously fail as well.
Without this patch the `PDFDocumentProxy.getMetadata` method, in the API, can thus fail which it *never* should and this will cause the viewer to not initialize all state as expected.
Fixes one of the documents in issue 14305.
*Please note:* This is similar to the method that existed prior to PR 3848, but the new method will *only* be used as a fallback when parsing of corrupt PDF documents.
The implementation in PR 14311 unfortunately turned out to be *way* too simplistic, as evident by the recently added test-files in issue 14303, since it may *cause* infinite loops in `PDFDocument.checkLastPage` for some corrupt PDF documents.[1]
To avoid this, the easiest solution that I could come up with was to fallback to eagerly parsing the *entire* /Pages-tree when the /Count-entry validation fails during document initialization.
Fixes *at least* two of the issues listed in issue 14303, namely the `poppler-395-0.pdf...` and `GHOSTSCRIPT-698804-1.pdf...` documents.
---
[1] The whole point of PR 14311 was obviously to *get rid of* infinte loops during document initialization, not to introduce any more of those.
This only applies to severely corrupt documents, where it's possible that the `Parser` throws when we try to access e.g. a /Kids-entry in the /Pages-tree.
Fixes two of the issues listed in issue 14303, namely the `poppler-742-0.pdf...` and `poppler-937-0.pdf...` documents.
*Please note:* While this patch on its own is sufficient to prevent the worker-thread from hanging, however in combination with PR 14311 these PDF documents will both load *and* render correctly.
Rather than focusing on the particular structure of these PDF documents, it seemed (at least to me) to make sense to try and prevent all circular references when fetching/looking-up data using the XRef table.
To avoid a solution that required tracking the references manually everywhere, the implementation settled on here instead handles that internally in the `XRef.fetch`-method. This should work, since that method *and* the `Parser`/`Lexer`-implementations are completely synchronous.
Note also that the existing `XRef`-caching, used for all data-types *except* Streams, should hopefully help to lessen the performance impact of these changes.
One *potential* problem with these changes could be certain *browser* exceptions, since those are generally not catchable in JavaScript code, however those would most likely "stop" worker-thread parsing anyway (at least I hope so).
Finally, note that I settled on returning dummy-data rather than throwing an exception. This was done to allow parsing, for the rest of the document, to continue such that *one* bad reference doesn't prevent an entire document from loading.
Fixes two of the issues listed in issue 14303, namely the `poppler-91414-0.zip-2.gz-53.pdf` and `poppler-91414-0.zip-2.gz-54.pdf` documents.
*This patch basically extends the approach from PR 10392, by also checking the last page.*
Currently, in e.g. the `Catalog.numPages`-getter, we're simply assuming that if the /Pages-tree has an *integer* /Count entry it must also be correct/valid.
As can be seen in the referenced PDF documents, that entry may be completely bogus which causes general parsing to breaking down elsewhere in the worker-thread (and hanging the browser).
Rather than hoping that the /Count entry is correct, similar to all other data found in PDF documents, we obviously need to validate it. This turns out to be a little less straightforward than one would like, since the only way to do this (as far as I know) is to parse the *entire* /Pages-tree and essentially counting the pages.
To avoid doing that for all documents, this patch tries to take a short-cut by checking if the last page (based on the /Count entry) can be successfully fetched. If so, we assume that the /Count entry is correct and use it as-is, otherwise we'll iterate through (potentially) the *entire* /Pages-tree to determine the number of pages.
Unfortunately these changes will have a number of *somewhat* negative side-effects, please see a possibly incomplete list below, however I cannot see a better way to address this bug.
- This will slow down initial loading/rendering of all documents, at least by some amount, since we now need to fetch/parse more of the /Pages-tree in order to be able to access the *last* page of the PDF documents.
- For poorly generated PDF documents, where the entire /Pages-tree only has *one* level, we'll unfortunately need to fetch/parse the *entire* /Pages-tree to get to the last page. While there's a cache to help reduce repeated data lookups, this will affect initial loading/rendering of *some* long PDF documents,
- This will affect the `disableAutoFetch = true` mode negatively, since we now need to fetch/parse more data during document initialization. While the `disableAutoFetch = true` mode should still be helpful in larger/longer PDF documents, for smaller ones the effect/usefulness may unfortunately be lost.
As one *small* additional bonus, we should now also be able to support opening PDF documents where the /Pages-tree /Count entry is completely invalid (e.g. contains a non-integer value).
Fixes two of the issues listed in issue 14303, namely the `poppler-67295-0.pdf` and `poppler-85140-0.pdf` documents.
For this particular PDF document, we have `/W [1 2 166666666666666666666666666]` which obviously makes no sense.
While this patch makes no attempt at actually validating the entries in the /W-array, we'll now simply abort all processing when the end of the PDF document has been reached (thus preventing hanging the browser).
Please note that this patch doesn't enable the PDF document to be loaded/rendered, but at least it fails "correctly" now.
Fixes one of the issues listed in issue 14303, namely the `REDHAT-1531897-0.pdf`document.
This bug was surprisingly difficult to track down, since it didn't just depend on range-requests being used but also on how quickly the document was loaded. To even be able to reproduce this locally, I had to use a very small `rangeChunkSize`-value (note the unit-test).
The cause of this bug is a bogus entry in the XRef-table, causing us to attempt to request data from *beyond* the actual document size and thus getting into an infinite loop.
Fixes *one* of the issues listed in issue 14303, namely the `PDFBOX-4352-0.pdf` document.
We were incorrectly using the transform in the pattern before it had been
adjusted causing the pattern to be misplaced relative to the page.
Fixes: ShowText-ShadingPattern.pdf (already in corpus)
Fixes: #8111Fixes: #9243
Subfrom nomin displays even though it's subform is set to <occur max=-1 min=0>
If we look through specs of XFA 3.3 : https://www.pdfa.org/norm-refs/XFA-3_3.pdf
- The min attribute is used when processing a form that contains data. Regardless of the data at least this number of instances is included. It is permissible to set this value to zero, in which case the container is entirely excluded if there is no data for it.
However, in our case it doesn't happen, because we let our empty dataNode get through. Though by setting a clause:
- eliminate unmatched data with occur min=0
we are checking our empty data and sending it to uselessNode array where at the end it gets removed;
Very short strings can narrowly miss the existing Bidi-detection threshold, leading to incorrect text-selection and copying behaviour.
In my testing, neither Adobe Reader or PDFium seem to handle copying "correctly" for this document. Hence it's not entirely clear to me that we actually want to fix this, since tweaking these heuristics can *obviously* cause regressions elsewhere (and our test coverage for RTL-text isn't exactly great).
It seems that issue 10301 was fixed by PR 13424, by combining the spans, however given that we don't have a lot of test coverage for RTL-text I figured that adding a simple reference test wouldn't hurt (rather than just closing the issue as WORKSFORME).
The old method of handling soft masks had a number of issues where the temporary
drawing canvas and the suspended main canvas could get out of sync
(e.g. mismatched save/restores or clip state) or we could end up compositing at
the wrong time. A good example of things getting out sync is the reduced test
case in #9017.
To fix this I've changed two big things:
1) Duplicate all the needed graphics state from the temporary canvas to the
suspended main canvas. This ensure the canvases stay in sync so that when we
switch back to the main canvas the graphics state stack is the same
(e.g. transforms, clip paths).
2) Immediately composite after each drawing operation. This ensures that if
there's an active clip region that we'll still be able to composite the correct
portions of the canvas. Note: This solution could be avoided by using
getImageData and putImageData since those ignore clipping region, but this is
very very slow. Note2: I also think the old way of only compositing at the end
of the soft mask is incorrect and can lead to wrong colors if drawing over the
same region, but in practice this doesn't seem to matter much.
Fixes: #5781Fixes: #5853Fixes: #7267Fixes: #7891Fixes: #8403Fixes: #8624Fixes: #12798Fixes: #13891Fixes: #9017 (reduced test case)
Fixes: https://bugzilla.mozilla.org/show_bug.cgi?id=1703683
- PR #13257 fixed a lot of issues but not all and this patch aims to fix almost all remaining issues.
- the idea in this new patch is to compare position of new glyph with the last position where a glyph has been drawn;
- no space are "drawn": it just moves the cursor but they aren't added in the chunk;
- so this way a space followed by a cursor move can be treated as only one space: it helps to merge all spaces into one.
- to make difference between real spaces and tracking ones, we used a factor of the space width (from the font)
- it was a pretty good idea in general but it fails with some fonts where space was too big:
- in Poppler, they're using a factor of the font size: this is an excellent idea (<= 0.1 * fontSize implies tracking space).
If a PDF included an embedded TrueType font whose preferred character
map (cmap) was in "format 2", the code would select that character map
and then refuse to read it because of an unsupported format, thus
causing the characters not to be rendered. This commit implements
support for format 2 as described at the link below.
https://developer.apple.com/fonts/TrueType-Reference-Manual/RM06/Chap6cmap.html
- it aims to fix#12721.
- Thanks to PR #14023, we've now the fieldObjects in the annotation layer so we can easily map fields names on their id if needed.
- Reset values in the storage, in the JS sandbox and in the visible html elements.
In the referenced bug, the embedded fonts contain custom CMap-data that only include strings. Note how for embedded composite TrueType fonts we're using the CMap-data when building the glyph mapping, and currently we end up with a completely empty map because the code expects only CID *numbers*.
Furthermore, just fixing the glyph mapping alone isn't sufficient to fully address the bug, since we also need to consider this "special" kind of CMap-data when looking up glyph widths.
In order to implement this, we utilize the existing `bidi` function to infer the text-direction of /T and /Contents entries. While this may not be perfect in cases where one PopupAnnotation mixes LTR and RTL languages, it should work well enough in most cases.
To avoid having to add *two new* properties in lots of annotations, supplementing the existing `title`/`contents`-properties, this patch instead re-factors the existing code such that the properties are replaced by Objects (containing `str` and `dir`).
*Please note:* In order avoid breaking existing third-party implementations, `GENERIC`-builds of the PDF.js library will still provide the old `title`/`contents`-properties on annotations returned by `PDFPageProxy.getAnnotations`.
Currently we only exclude /Encoding entries that also contains a /Differences array, which is the cause of the text-selection problem in the referenced issue.
In order to address this we'll now also exclude /Encoding entries that contain one of the predefined *named* encodings, and no longer require that it also contains a /Differences array.
*Please note:* This patch cases a small "regression" in the `bug1130815-text` test-case, however this is actually an improvement when compared with Adobe Reader and PDFium (in Google Chrome).
*Please note:* All of this feels very handwavy, but at least it passes all tests locally. Hopefully we have enough tests for this part of the font code.
For non-embedded composite standard fonts with an "incomplete" /CIDToGIDMap, we'll now fallback to an *explicitly defined* /ToUnicode map even when that one happens to be an /Identity-H or /Identity-V map.
The `Font.fallbackToSystemFont` method is unfortunately getting more and more special-cases, however that might be unavoidable given all the weird non-embedded fonts found in the wild :-(
While some of the output looks worse to my eye, this behavior more
closely matches what I see when I open the PDFs in Adobe acrobat.
Fixes: #4706, #9713, #8245, #1344
In the PDF document some of the glyphs have bogus `differences`-entries[1] that cannot be resolved to valid glyph names, thus causing the glyph mapping to fail.
My initial idea was to use a similar approach as in the `PartialEvaluator._simpleFontToUnicode`-method, to extract the charCodes from those entries, however it turned out that that didn't actually help in this case (the mapping was still wrong).
To fix this I'm thus proposing that we fallback to the /ToUnicode map when no other useable data exists (e.g. no post-table), since it *hopefully* shouldn't make things any worse than leaving parts of the glyph map empty (which currently happens).
---
[1] As can be seem below, some of the entries are completely normal while others are non-standard:
```
Differences (array)
0 = 65
1 = /g5167
2 = /space
3 = /g11927
4 = /g17737
5 = /g11540
6 = /g2180
7 = /K
8 = /P
9 = /two
10 = /zero
11 = /one
12 = /five
13 = /four
14 = /g6932
15 = /g7246
16 = /g1691
17 = /g2343
18 = /g14792
19 = /g3325
20 = /g4280
21 = /g20383
22 = /g18166
23 = /g16988
24 = /g17943
25 = /g19223
26 = /g10830
27 = 97
28 = /g982
29 = /g1226
30 = /g5059
31 = /g2677
32 = /g1042
33 = /g11568
34 = /L
35 = /three
36 = /seven
37 = /g2364
38 = /g12063
39 = /g5356
40 = /g2173
41 = /g17877
42 = /g7273
43 = /g7647
44 = /g7224
45 = /g19327
46 = /g5054
47 = /g2342
48 = /g10136
49 = /g6856
50 = /g13381
51 = /g7257
52 = /g12093
53 = /g2359
```
Currently, in the `PartialEvaluator`, we only support Optional Content in Form-/XObjects. Hence this patch adds support for Image-/XObjects as well, which looks like a simple oversight in PR 12095 since the canvas-implementation already contains the necessary code to support this.
Bug 1721218 has a shading pattern that was used thousands of times.
To improve performance of this PDF:
- add a cache for patterns in the evaluator and only send the IR form once
to the main thread (this also makes caching in canvas easier)
- cache the created canvas radial/axial patterns
- for shading fill radial/axial use the pattern directly instead of creating temporary
canvas
Previously, when we filled image masks we didn't copy over the current transformation,
this caused patterns to be misaligned when painted. Now we create a temporary
canvas with the mask and have the transform copied over and offset it relative to
where the mask would be painted. We also weren't properly offsetting tiling patterns.
This isn't usually noticeable since patters repeat, but in the case of #13561 the pattern
is only drawn once and has to be in the correct position to line up with the mask image.
These fixes broke #11473, but highlighted that we were drawing that correctly by
accident and not correctly handling negative bounding boxes on tiling patterns.
Fixes#6297, #13561, #13441
Partially fixes#1344 (still blurry but boxes are in correct position now)
- when binding (after parsing) we get a map between some template nodes and some data nodes;
- so set user data in input handlers in using data node uids in the annotation storage;
- to save the form, just put the value we have in the storage in the correct data nodes, serialize the xml as a string and then write the string at the end of the pdf using src/core/writer.js;
- fix few bugs around data bindings:
- the "Off" issue in Bug 1716980.
Note, this only really fixes Radial/Axial shading patterns with masks.
I'm guessing tiling patterns and mesh patterns would also be broken
if applied like the test pdf. Hopefully I'll have some time to make
test cases for the other shadings.
Fixes#13372
For Type3 fonts where the /CharProcs-streams of the individual glyph starts with a `d1` operator, we can use that to build a fallback bounding box for the font and thus improve text-selection in some cases.
For HighlightAnnotations with a built-in appearance stream, we still rely on it to specify the opacity correctly via a suitable blend mode. However, if the Annotation-drawing operators are placed *within* a /XObject of the /Form-type, the /ExtGState won't apply to the final rendering and the result is that the highlighting obscures the underlying text.
The more *correct* and general solution would likely be to somehow modify the implementation in `src/display/canvas.js`, to special-case handling of /Form-type /XObjects when rendering Annotations. Since we can very easily work-around this problem for now by using the "no appearance stream" code-path, doing *something* here ought to be preferable.
This patch is (obviously) merely a work-around, but given that the referenced issue is (as far as I know) the first case we've seen of this problem a simple solution will hopefully suffice for now.
This fixes the colours, by respecting the strokeAlpha/fillAlpha-values, for a couple of Annotations in the PDF document from issue 13447.[1]
---
[1] Some of the annotations still won't render at all, when compared with Adobe Reader, but that could/should probably be handled separately.
