[api-minor] Include and use the 14 standard font files. #12726

brendandahl · 2020-12-11T01:36:13Z

Fixes #4244 (at least for the standard font issue).

I need to do some more testing to see if this impacts performance for the browser use case. I'm also considering disabling this feature for browsers so we use the builtin fonts. Though, we may want to use the various symbol fonts since those seem to vary per platform.

brendandahl · 2020-12-11T01:36:25Z

/botio test

pdfjsbot · 2020-12-11T01:36:26Z

From: Bot.io (Linux m4)

Received

Command cmd_test from @brendandahl received. Current queue size: 0

Live output at: http://54.67.70.0:8877/5b052292e34e465/output.txt

pdfjsbot · 2020-12-11T01:36:26Z

From: Bot.io (Windows)

Received

Command cmd_test from @brendandahl received. Current queue size: 0

Live output at: http://3.101.106.178:8877/9cd5c996cf5da70/output.txt

pdfjsbot · 2020-12-11T02:02:55Z

From: Bot.io (Linux m4)

Failed

Full output at http://54.67.70.0:8877/5b052292e34e465/output.txt

Total script time: 26.47 mins

Font tests: Passed
Unit tests: FAILED
Integration Tests: Passed
Regression tests: FAILED

Image differences available at: http://54.67.70.0:8877/5b052292e34e465/reftest-analyzer.html#web=eq.log

pdfjsbot · 2020-12-11T02:05:08Z

From: Bot.io (Windows)

Failed

Full output at http://3.101.106.178:8877/9cd5c996cf5da70/output.txt

Total script time: 28.68 mins

Font tests: Passed
Unit tests: FAILED
Integration Tests: Passed
Regression tests: FAILED

Image differences available at: http://3.101.106.178:8877/9cd5c996cf5da70/reftest-analyzer.html#web=eq.log

Snuffleupagus

Overall this seems like a very reasonable approach, and the re-use between the CMap/StandardFont factories is really nice.
Besides a few inline comments, based on a quick look at the code, I've got one "bigger picture" suggestion below.

I'm also considering disabling this feature for browsers so we use the builtin fonts. Though, we may want to use the various symbol fonts since those seem to vary per platform.

My general train of thought here (ignoring the symbol fonts for now), is that we'd only fetch/use the standard font files when actually necessary. In practice, that'd mean:

When the disableFontFace API-parameter is true.
When handling Patterns, see e.g.

pdf.js/src/core/evaluator.js

Line 855 in 00b4f86

state.fillColorSpace.name === "Pattern" ||
When failing to load a font and we thus fallback, see e.g.

pdf.js/src/core/evaluator.js

Line 856 in 00b4f86

font.disableFontFace ||

Edit: Thinking about my suggestion a bit more, it seems that conditional loading will probably be somewhat difficult to implement well given the existing font-parsing code and the multitude of cases to consider. However, with caching implemented in the worker (note the inline comment), I'm not sure how much of a problem always loading these font files would really be!?

(Back when I added worker-thread caching of CMap-files, overall performance improved significantly for affected documents.)

src/core/evaluator.js

Snuffleupagus · 2020-12-11T15:30:17Z

src/core/evaluator.js

+        let file = null;
+        if (standardFontName) {
+          const data = await this.fetchStandardFontData(standardFontName);
+          file = new Stream(data);


Please let fetchStandardFontData return the data as a Stream, rather than doing that manually at the call-sites.

Snuffleupagus · 2020-12-11T15:36:38Z

src/core/fonts.js

@@ -243,6 +243,12 @@ function getFontType(type, subtype) {
  }
 }

+function getStandardFontName(name) {
+  const fontName = name.replace(/[,_]/g, "-").replace(/\s/g, "");


Since you're using the same regular expression at

pdf.js/src/core/fonts.js

Line 1319 in 00b4f86

let fontName = name.replace(/[,_]/g, "-").replace(/\s/g, "");

can you please extract the regexp-part to a new constant, used by both call-sites, to avoid these getting out of sync?

src/display/api.js

Snuffleupagus · 2020-12-11T15:42:17Z

src/display/display_utils.js

-        return { cMapData, compressionType };
-      });
-    }
+    return fetchData(url, this.isCompressed).then(data => {


Nit:

Suggested change

return fetchData(url, this.isCompressed).then(data => {

return fetchData(url, /* asTypedArray = */ this.isCompressed).then(data => {

Snuffleupagus · 2020-12-11T15:42:48Z

src/display/display_utils.js

    });
  }
 }

+class DOMStandardFontDataFactory extends BaseStandardFontDataFactory {
+  _fetchData(url) {
+    return fetchData(url, true);


Nit:

Suggested change

return fetchData(url, true);

return fetchData(url, /* asTypedArray = */ true);

Snuffleupagus · 2020-12-11T15:45:10Z

web/app_options.js

@@ -178,6 +178,14 @@ const defaultOptions = {
        : "../web/cmaps/",
    kind: OptionKind.API,
  },
+  standardFontDataUrl: {


Please move this further down, since the parameters for each OptionKind should be sorted alphabetically.

Snuffleupagus · 2020-12-11T15:56:08Z

src/core/evaluator.js

+    const standardFontNameToFileName = {
+      Courier: "FoxitFixed",
+      "Courier-Bold": "FoxitFixedBold",
+      "Courier-BoldOblique": "FoxitFixedBoldItalic",
+      "Courier-Oblique": "FoxitFixedItalic",
+      Helvetica: "FoxitSans",
+      "Helvetica-Bold": "FoxitSansBold",
+      "Helvetica-BoldOblique": "FoxitSansBoldItalic",
+      "Helvetica-Oblique": "FoxitSansItalic",
+      "Times-Roman": "FoxitSerif",
+      "Times-Bold": "FoxitSerifBold",
+      "Times-BoldItalic": "FoxitSerifBoldItalic",
+      "Times-Italic": "FoxitSerifItalic",
+      Symbol: "FoxitSymbol",
+      ZapfDingbats: "FoxitDingbats",
+    };


It might not hurt to move this to https://github.com/mozilla/pdf.js/blob/master/src/core/standard_fonts.js instead!?

Snuffleupagus · 2021-01-10T11:02:07Z

Given the sheer number of long-standing issues/bugs that this PR would solve, this seems like something that would be really nice to try and fix-up and land sooner rather than later :-)
@brendandahl Do you have time to finish this, such that we can land the PR?

Also, we're probably overdue for a new PDF.js release at this point; however I'd suggest that we hold off on that a little while longer since this PR would be really nice to include in the next release. /cc @timvandermeij

timvandermeij · 2021-01-10T11:57:37Z

I agree, this would be a really nice one to include.

brendandahl · 2021-01-22T22:54:47Z

Still waiting on some Firefox talos runs for windows, but locally this patch is much slower loading the first page of unix01.pdf in the pdfpaint benchmark. 20 runs median 238.43ms vs 332.51ms

brendandahl · 2021-01-23T04:37:48Z

Windows talos runs also show a slow down:
base: 257
new: 313
https://treeherder.mozilla.org/perfherder/compare?originalProject=try&originalRevision=2f45f4bcccb0fc60c48180aaaa432e5f1d96424d&newProject=try&newRevision=5cac140a3074ed804b1d72413384f2f065ed83d7&framework=1&showOnlyImportant=1

Snuffleupagus · 2021-01-23T11:13:34Z

Windows talos runs also show a slow down:
base: 257
new: 313
https://treeherder.mozilla.org/perfherder/compare?originalProject=try&originalRevision=2f45f4bcccb0fc60c48180aaaa432e5f1d96424d&newProject=try&newRevision=5cac140a3074ed804b1d72413384f2f065ed83d7&framework=1&showOnlyImportant=1

I suppose that it's not entirely surprising if some slowdown is observed, since we're now having to actually parse font data that we previously didn't have to. The question is obviously how large of a regression we can accept!?

Looking briefly at the tested patch, it appears that it doesn't actually implement worker-thread caching of the standard font data (something that I mentioned as a potential problem above). If that's indeed the case, I'd not be entirely surprised if that at least in part contributes to the observed regression.

What would be very interesting to know, is more precisely where the regression comes from:

If it's only caused by src/core/cff_parser.js being used a lot more, then a regression would seem entirely acceptable since we obviously need to parse more font data with this patch.
If, on the other hand, a significant part of the regression is related to the additional data loading itself then I'd suggest that worker-thread caching might help. A related idea could perhaps be to pre-load the font data early during document loading, to cut down on the waiting time during the actual font-parsing!?

All-in-all, I suppose that it's going to take more work to get this landed than I'd initially hoped.

Snuffleupagus · 2021-01-23T12:45:56Z

Having played around with this patch locally, focusing on the actual standard font data loading (using console.time/timeEnd), it seems that a very significant part of the regression comes from just waiting for the font data to become available in the worker.

Based on that observation I have one idea, however it's not a great one: What if we simply bundled the standard font data into the pdf.worker.js file, such that it's synchronously available?
Even suggesting this feel all kinds of wrong, but I'm still wondering it this would be an acceptable trade-off here?

It's difficult to not compare with the CMaps, but I do believe that this situation is slightly different here:

A very large number of existing PDF files use non-embedded standard fonts, whereas CMaps are much less common in practice (e.g. mostly non-Latin documents). Hence it's going to be very common to have to access the standard font data.
The standard font data files included in this patch take up 266 616 bytes, whereas the CMap files are 1 167 756 bytes in total. Obviously the pdf.worker.js is already large, but that kind of size increase doesn't seem completely unacceptable in this case.
We've seen again and again how people report bugs, in custom PDF.js implementations, with broken font rendering where the culprit is that the CMap-parameters weren't provided when calling getDocument[1]; bundling the standard font data into pdf.worker.js would prevent similar issues there.
Given how many duplicate issues/bugs we've seen regarding missing standard fonts in PDF.js, it seems very unlikely that users would not want to use this feature anyway.

[1] Many users seem to just ignore the errors/warnings, mentioning the problem, printed in the console.

brendandahl · 2021-01-28T17:39:09Z

I also tried fetching the font data from the worker. It was a bit faster but not much.

I'd be curious to try directly embedding the files in the worker, but 266,616 bytes seems like a big increase for something that is not really needed in the common case.

Looking briefly at the tested patch, it appears that it doesn't actually implement worker-thread caching of the standard font data (something that I mentioned as a potential problem above). If that's indeed the case, I'd not be entirely surprised if that at least in part contributes to the observed regression.

For the talos test, the web page is reloaded so the cache wouldn't help there. Also, I only see the font data being loaded once right now since we cache fonts, so I'm not sure this would help in the re-render case either.

I'm planning to put this patch on the back burner though as we have some higher priority stuff for firefox (XFA and tagged PDFs). I'll push up my changes with comments addressed, then we could add a pref to have it disabled by default for the time being.

Snuffleupagus · 2021-01-28T17:48:30Z

I'd be curious to try directly embedding the files in the worker, but 266,616 bytes seems like a big increase for something that is not really needed in the common case.

One more idea here: How about generating a separate bundle with the standard font data, and then using importScripts to load it into the worker-thread when needed?

Edit: Or how about simply generating one large font data file, similar to but a lot simpler than the bcmap ones, such that we only need to fetch a single (larger) file once rather than do multiple requests for the individual standard font data files?
Based on some very quick testing with pdfBug=Stats, it seems that loading only one file would help reduce the overhead.

Snuffleupagus · 2021-01-31T18:59:55Z

I've tried implementing the second idea in #12726 (comment), i.e. creating a bundle with all of the standard font data such that we only need to fetch a single file.
Despite the bundle being fairly large (~260 kb), this is a significant performance improvement for the unix01.pdf document (~33 percent faster than this PR). Obviously it's still not as fast as the master branch, but the remaining slowdown is unavoidable since there's more parsing happening in src/core/fonts.js and src/core/cff_parser.js.

Please find my patch at master...Snuffleupagus:standard-fonts

@brendandahl This is probably the best that we can do, without inlining the standard font data in the pdf.worker.js file; could we perhaps consider moving forward with our combined patches here?

Snuffleupagus · 2021-02-02T13:15:32Z

[...] as we have some higher priority stuff for firefox (XFA and tagged PDFs).

I obviously don't question the value of either of those features.
However, I'm willing to bet that including standard font data in PDF.js will fix a lot more things than the above features combined :-)

Snuffleupagus · 2021-02-19T11:57:06Z

@brendandahl Ping, how about my idea/implementation mentioned in #12726 (comment)?

Given that it's going to be impossible to add this functionality without any sort of performance overhead, since we now need to parse more font data in affected cases, the only thing that we can realistically do here is to limit the necessary data fetching/loading by only having one (larger) font data bundle.

brendandahl · 2021-02-24T21:24:19Z

I've tried implementing the second idea in #12726 (comment), i.e. creating a bundle with all of the standard font data such that we only need to fetch a single file.
Despite the bundle being fairly large (~260 kb), this is a significant performance improvement for the unix01.pdf document (~33 percent faster than this PR). Obviously it's still not as fast as the master branch, but the remaining slowdown is unavoidable since there's more parsing happening in src/core/fonts.js and src/core/cff_parser.js.

Please find my patch at master...Snuffleupagus:standard-fonts

@brendandahl This is probably the best that we can do, without inlining the standard font data in the pdf.worker.js file; could we perhaps consider moving forward with our combined patches here?

I imagine that will slow down the case where there's only one font used (especially for the generic version of the viewer)?

Another thought, we could try to translate the fonts so they don't need to go through the full font parser. I imagine this would be quite a bit of worker though. It'd probably be good to profile some more though, and see if the font translation is slowing it down or just loading the font in the browser is the slow part.

Snuffleupagus · 2021-02-25T14:49:40Z

I imagine that will slow down the case where there's only one font used (especially for the generic version of the viewer)?

At least locally, the difference in loading time is quite small between loading one font file and loading the (larger) bundle. It seems to me that the overhead related to simply opening the connection, fetching the data, and finally transferring it to the worker-thread is what dominates overall.
However, it would be quite simple to by default only use the bundle in MOZCENTRAL builds, but use the individual files for the GENERIC build. I purposely modified the relevant code such that both "file formats" would still be supported :-)

Another thought, we could try to translate the fonts so they don't need to go through the full font parser. I imagine this would be quite a bit of worker though.

I've also had that idea, but thinking about the practical implications of that made me quickly abandon it. Besides the overall complexity, we'd also need to add a way to automatically re-build the translated fonts for any code-changes.

[...] and see if the font translation is slowing it down or just loading the font in the browser is the slow part.

With the bundle approach, it's mostly the actual font conversion code that contributes to the "regression" here.

However, I do believe that the comparison is perhaps not entirely fair here:
In an ideal world every PDF document would simply embed all of its fonts, and we'd thus always have to run our font conversion code.
Hence it may be slightly more fair to compare the performance for a PDF document which has embedded standard fonts, but instead forcibly load "our" standard font data!?

Snuffleupagus · 2021-03-02T23:16:51Z

Another thing that I think this patch would also improve is rendering of standard fonts on Linux, since we've seen a few bugs/issues reported where non-embedded standard fonts render badly on Linux because of poor font substitution.

Some examples include https://bugzilla.mozilla.org/show_bug.cgi?id=1695727 and issue #11840

Snuffleupagus · 2021-03-19T10:16:21Z

I'd be curious to try directly embedding the files in the worker, but 266,616 bytes seems like a big increase for something that is not really needed in the common case.

Well, with recent development I don't really think that argument holds any more :-)

The XFA functionality, which just landed, increased the size of the built pdf.worker.js file alone by 214 903 bytes in the MOZCENTRAL build. In my experience, XFA is quite uncommon whereas every other PDF file would benefit from us bundling standard fonts.
Hence I think that if we're willing to accept such a large size increase for a niche feature such as XFA, it really ought to be acceptable for non-embedded standard fonts given just how common they are.

rtestard · 2021-03-19T16:28:17Z

The XFA functionality, which just landed, increased the size of the built pdf.worker.js file alone by 214 903 bytes in the MOZCENTRAL build. In my experience, XFA is quite uncommon whereas every other PDF file would benefit from us bundling standard fonts.
Hence I think that if we're willing to accept such a large size increase for a niche feature such as XFA, it really ought to be acceptable for non-embedded standard fonts given just how common they are.

XFA is a lot less frequent than Acroform but we see about a million unique monthly active users encountering them in our desktop telemetry. I know at least of the Canadian Gov website using them. This is a significant number of users who will benefit from this change.

Snuffleupagus · 2021-03-19T16:54:02Z

XFA is a lot less frequent than Acroform but we see about a million unique monthly active users encountering them in our desktop telemetry.

In all fairness, then we should also measure how often users encounter non-embedded standard fonts in PDF documents; I'd be extremely surprised if that number isn't at least one order of magnitude larger :-)

marco-c · 2021-03-25T16:47:40Z

I know at least of the Canadian Gov website using them

There are some XFA forms on the UK and USA gov websites too.

pdfjsbot · 2021-06-07T20:05:55Z

From: Bot.io (Linux m4)

Received

Command cmd_test from @brendandahl received. Current queue size: 0

Live output at: http://54.67.70.0:8877/7e9df84dd3886ac/output.txt

pdfjsbot · 2021-06-07T20:05:55Z

From: Bot.io (Windows)

Received

Command cmd_test from @brendandahl received. Current queue size: 0

Live output at: http://3.101.106.178:8877/bc88cc4cd488960/output.txt

pdfjsbot · 2021-06-07T20:32:07Z

From: Bot.io (Linux m4)

Failed

Full output at http://54.67.70.0:8877/7e9df84dd3886ac/output.txt

Total script time: 26.19 mins

Font tests: Passed
Unit tests: Passed
Integration Tests: Passed
Regression tests: FAILED

Image differences available at: http://54.67.70.0:8877/7e9df84dd3886ac/reftest-analyzer.html#web=eq.log

pdfjsbot · 2021-06-07T20:35:34Z

From: Bot.io (Windows)

Failed

Full output at http://3.101.106.178:8877/bc88cc4cd488960/output.txt

Total script time: 29.65 mins

Font tests: Passed
Unit tests: Passed
Integration Tests: Passed
Regression tests: FAILED

Image differences available at: http://3.101.106.178:8877/bc88cc4cd488960/reftest-analyzer.html#web=eq.log

Snuffleupagus · 2021-06-08T08:09:47Z

/botio makeref

pdfjsbot · 2021-06-08T08:09:48Z

From: Bot.io (Linux m4)

Received

Command cmd_makeref from @Snuffleupagus received. Current queue size: 0

Live output at: http://54.67.70.0:8877/cdfe0a1ad516c1b/output.txt

pdfjsbot · 2021-06-08T08:09:48Z

From: Bot.io (Windows)

Received

Command cmd_makeref from @Snuffleupagus received. Current queue size: 0

Live output at: http://3.101.106.178:8877/029042f654db4da/output.txt

pdfjsbot · 2021-06-08T08:32:34Z

From: Bot.io (Linux m4)

Success

Full output at http://54.67.70.0:8877/cdfe0a1ad516c1b/output.txt

Total script time: 22.76 mins

Lint: Passed
Make references: Passed
Check references: Passed

pdfjsbot · 2021-06-08T08:36:18Z

From: Bot.io (Windows)

Success

Full output at http://3.101.106.178:8877/029042f654db4da/output.txt

Total script time: 26.50 mins

Lint: Passed
Make references: Passed
Check references: Passed

- a lot of xfa files are using Myriad pro or Arial fonts without embedding them and some containers have some dimensions based on those font metrics. So not having the exact same font leads to a wrong display. - since it's pretty hard to find a replacement font with the exact same metrics, this patch gives the possibility to read glyf table, rescale each glyph and then write a new table. - so once PR mozilla#12726 is merged we could rescale for example Helvetica to replace Myriad Pro.

kmcclellan · 2021-06-08T13:27:16Z

Could we get a new build of pdfjs-dist to npm please?

Snuffleupagus · 2021-06-08T13:35:58Z

Could we get a new build of pdfjs-dist to npm please?

Not immediately, since this just landed and we (obviously) need some time to handle any possible regressions and there's also various follow-up work that should happen first. (Also, the last release was less than two weeks ago.)

Snipeye · 2021-06-09T03:57:59Z

I don't github (verb) very often, so I don't know if this is correct/appropriate (or if I'm about to ping a million people, in which case I am deeply sorry) but I just read through this and wanted to add my "thank you 1e6" for the effort to get this in; at least in my case it dramatically increases the usability.

- a lot of xfa files are using Myriad pro or Arial fonts without embedding them and some containers have some dimensions based on those font metrics. So not having the exact same font leads to a wrong display. - since it's pretty hard to find a replacement font with the exact same metrics, this patch gives the possibility to read glyf table, rescale each glyph and then write a new table. - so once PR mozilla#12726 is merged we could rescale for example Helvetica to replace Myriad Pro.

ceztko · 2021-11-19T09:23:08Z

Sorry for the late comment. I just wanted to say that the fonts taken from PDFium are possibly misnamed, or their format is not immediately clear. pfb extension seems to be mostly taken for Printer Font Binary, which is basically the encrypted part of a pfa font. pfb can't be used alone but must be coupled with a pfm file that just contains the metrics. The fonts taken from PDFium are neither canonical pfb or pfa, but they seem to be a raw OpenType compatible CFF representation, and I can confirm they are opened by the CFF driver in Freetype. I don't know if there's a better extension for these. At least their actual format could be better documented.

- a lot of xfa files are using Myriad pro or Arial fonts without embedding them and some containers have some dimensions based on those font metrics. So not having the exact same font leads to a wrong display. - since it's pretty hard to find a replacement font with the exact same metrics, this patch gives the possibility to read glyf table, rescale each glyph and then write a new table. - so once PR mozilla#12726 is merged we could rescale for example Helvetica to replace Myriad Pro.

brendandahl changed the title ~~Include and use the 14 standard fonts files.~~ Include and use the 14 standard font files. Dec 11, 2020

Snuffleupagus requested changes Dec 11, 2020

View reviewed changes

Snuffleupagus reviewed Dec 11, 2020

View reviewed changes

timvandermeij added the font-conversion label Dec 11, 2020

Snuffleupagus mentioned this pull request Jan 11, 2021

Move 2.6.347 to release and add new release #12851

Closed

brendandahl force-pushed the standard-fonts branch from 34a28ab to bc41373 Compare January 28, 2021 19:46

timvandermeij mentioned this pull request Feb 25, 2021

Error when loading PDF that uses system fonts #4244

Closed

Snuffleupagus approved these changes Jun 8, 2021

View reviewed changes

Snuffleupagus merged commit e7dc822 into mozilla:master Jun 8, 2021

This was referenced Jun 8, 2021

[api-minor] Fetch binary CMap data in the worker-thread, when useWorkerFetch is set #13514

Merged

Cache the "raw" standard font data in the worker-thread (PR 12726 follow-up) #13515

Merged

Snuffleupagus mentioned this pull request Jun 8, 2021

Always use standard font data, with disableFontFace set in the API (PR 12726 follow-up) #13516

Merged

kmcclellan mentioned this pull request Jun 18, 2021

TrueType standard fonts #13585

Closed

Snuffleupagus mentioned this pull request Jun 19, 2021

Embedded fonts are not rendered in node when rendering to canvas #7929

Closed

Snuffleupagus mentioned this pull request Jun 29, 2021

Can not use standardFontDataUrl #13646

Closed

Snuffleupagus mentioned this pull request Aug 18, 2021

Font Not loading Correctly #13907

Closed

Snuffleupagus mentioned this pull request Nov 7, 2021

Determining when to use disableFontFace #11311

Closed

This was referenced Apr 26, 2022

Replace most Foxit fonts with Liberation fonts #14843

Open

Use a local font when the font isn't embedded (bug 1766039) #14844

Closed

Snuffleupagus mentioned this pull request Apr 17, 2023

PDF text rendering modes (Tr operator) that clip do not work #16302

Open

Snuffleupagus mentioned this pull request May 3, 2023

Remove the unused isStandardFont font-property (PR 15880 follow-up) #16381

Merged

	return fetchData(url, this.isCompressed).then(data => {
	return fetchData(url, /* asTypedArray = */ this.isCompressed).then(data => {

	return fetchData(url, true);
	return fetchData(url, /* asTypedArray = */ true);

[api-minor] Include and use the 14 standard font files. #12726

[api-minor] Include and use the 14 standard font files. #12726

Conversation

brendandahl commented Dec 11, 2020

brendandahl commented Dec 11, 2020

pdfjsbot commented Dec 11, 2020

From: Bot.io (Linux m4)

Received

pdfjsbot commented Dec 11, 2020

From: Bot.io (Windows)

Received

pdfjsbot commented Dec 11, 2020

From: Bot.io (Linux m4)

Failed

pdfjsbot commented Dec 11, 2020

From: Bot.io (Windows)

Failed

Snuffleupagus left a comment • edited Loading

Choose a reason for hiding this comment

Snuffleupagus Dec 11, 2020

Choose a reason for hiding this comment

Snuffleupagus Dec 11, 2020

Choose a reason for hiding this comment

Snuffleupagus Dec 11, 2020

Choose a reason for hiding this comment

Snuffleupagus Dec 11, 2020

Choose a reason for hiding this comment

Snuffleupagus Dec 11, 2020

Choose a reason for hiding this comment

Snuffleupagus Dec 11, 2020

Choose a reason for hiding this comment

Snuffleupagus commented Jan 10, 2021

timvandermeij commented Jan 10, 2021

brendandahl commented Jan 22, 2021

brendandahl commented Jan 23, 2021

Snuffleupagus commented Jan 23, 2021

Snuffleupagus commented Jan 23, 2021 • edited Loading

brendandahl commented Jan 28, 2021

Snuffleupagus commented Jan 28, 2021 • edited Loading

Snuffleupagus commented Jan 31, 2021 • edited Loading

Snuffleupagus commented Feb 2, 2021

Snuffleupagus commented Feb 19, 2021

brendandahl commented Feb 24, 2021

Snuffleupagus commented Feb 25, 2021 • edited Loading

Snuffleupagus commented Mar 2, 2021 • edited Loading

Snuffleupagus commented Mar 19, 2021 • edited Loading

rtestard commented Mar 19, 2021

Snuffleupagus commented Mar 19, 2021

marco-c commented Mar 25, 2021 • edited Loading

pdfjsbot commented Jun 7, 2021

From: Bot.io (Linux m4)

Received

pdfjsbot commented Jun 7, 2021

From: Bot.io (Windows)

Received

pdfjsbot commented Jun 7, 2021

From: Bot.io (Linux m4)

Failed

pdfjsbot commented Jun 7, 2021

From: Bot.io (Windows)

Failed

Snuffleupagus commented Jun 8, 2021

pdfjsbot commented Jun 8, 2021

From: Bot.io (Linux m4)

Received

pdfjsbot commented Jun 8, 2021

From: Bot.io (Windows)

Received

pdfjsbot commented Jun 8, 2021

From: Bot.io (Linux m4)

Success

pdfjsbot commented Jun 8, 2021

From: Bot.io (Windows)

Success

kmcclellan commented Jun 8, 2021

Snuffleupagus commented Jun 8, 2021

Snipeye commented Jun 9, 2021

ceztko commented Nov 19, 2021 • edited Loading

Snuffleupagus left a comment •

edited

Loading

Snuffleupagus commented Jan 23, 2021 •

edited

Loading

Snuffleupagus commented Jan 28, 2021 •

edited

Loading

Snuffleupagus commented Jan 31, 2021 •

edited

Loading

Snuffleupagus commented Feb 25, 2021 •

edited

Loading

Snuffleupagus commented Mar 2, 2021 •

edited

Loading

Snuffleupagus commented Mar 19, 2021 •

edited

Loading

marco-c commented Mar 25, 2021 •

edited

Loading

ceztko commented Nov 19, 2021 •

edited

Loading