2034406 - [skpdf] PDFs generated from pdf.js (via MozPrintCallback) are missing selectable text

Status: VERIFIED FIXED

(Core :: Printing: Output, defect, P3)

Description

[Calixte Denizet (:calixte)]

Attachment: tracemonkey.pdf

STR:

1. Set about:config pref print.experimental.skpdf to true
2. Print, and choose the Save to PDF backend.
3. Open the resulting file in Firefox
4. Try to select any text in the document.

ACTUAL RESULTS:
The text is unselectable.

EXPECTED RESULTS:
Text should be selectable.

Each page is rendered as an image.

Comment 1

[Lee Salzman [:lsalzman]]

Jamie, Emilio, can you take a look?

Comment 2

[Emilio Cobos Álvarez [:emilio]]

Happens on Linux as well afaict.

Comment 3

[Emilio Cobos Álvarez [:emilio]]

Calixte, how does pdf.js draw the text for printing? Just .drawText from the mozPrintCallback?

Comment 4

[Calixte Denizet (:calixte)]

(In reply to Emilio Cobos Álvarez [:emilio] from comment #3)

Calixte, how does pdf.js draw the text for printing? Just .drawText from the mozPrintCallback?

Yes just drawText.
I don't know if it matters but we call drawText for each char.

Comment 5

[Emilio Cobos Álvarez [:emilio]]

Yeah so I think I understand what's going on. What happens is that for MozPrintCallback, we call CreateSimilarDrawTarget for each <canvas> element (for each page, effectively):

• https://searchfox.org/firefox-main/rev/33682acc1fa0db34ac826b143048db28ee9f16a6/layout/generic/nsPageSequenceFrame.cpp#670-671

That basically causes it to get a raster surface here:

• https://searchfox.org/firefox-main/rev/33682acc1fa0db34ac826b143048db28ee9f16a6/gfx/2d/DrawTargetSkia.cpp#1678-1686

Lee, I'm not sure what the best way to go about this is, here's what I've tried:

I first tried the obvious "implement CreateSimilarDrawTarget without a raster surface". I don't see a particularly trivial way of doing that tho. SkPDFDocument::makeSurface just returns a raster surface, and reusing the SkCanvas seems like asking for trouble. Of course that works in this particular case because we only have one canvas per page, but still.

I tried to use the recording surface stuff with something like the diff below:

diff --git a/layout/generic/nsPageSequenceFrame.cpp b/layout/generic/nsPageSequenceFrame.cpp
index 7d1e0c27bff7..3cc894d849a3 100644
--- a/layout/generic/nsPageSequenceFrame.cpp
+++ b/layout/generic/nsPageSequenceFrame.cpp
@@ -12,6 +12,8 @@
 #include "mozilla/PrintedSheetFrame.h"
 #include "mozilla/StaticPresData.h"
 #include "mozilla/dom/HTMLCanvasElement.h"
+#include "mozilla/gfx/2D.h"
+#include "mozilla/gfx/DrawEventRecorder.h"
 #include "mozilla/gfx/Point.h"
 #include "mozilla/intl/AppDateTimeFormat.h"
 #include "nsCOMPtr.h"
@@ -659,16 +661,18 @@ nsresult nsPageSequenceFrame::PrePrintNextSheet(nsITimerCallback* aCallback,
       UniquePtr<gfxContext> renderingContext = dc->CreateRenderingContext();
       NS_ENSURE_TRUE(renderingContext, NS_ERROR_OUT_OF_MEMORY);

-      DrawTarget* drawTarget = renderingContext->GetDrawTarget();
-      if (NS_WARN_IF(!drawTarget)) {
+      DrawTarget* referenceDt = renderingContext->GetDrawTarget();
+      if (NS_WARN_IF(!referenceDt)) {
         return NS_ERROR_FAILURE;
       }

       for (HTMLCanvasElement* canvas : Reversed(mCurrentCanvasList)) {
         CSSIntSize size = canvas->GetSize();
-
-        RefPtr<DrawTarget> canvasTarget = drawTarget->CreateSimilarDrawTarget(
-            size.ToUnknownSize(), drawTarget->GetFormat());
+        RefPtr recorder = MakeAndAddRef<gfx::DrawEventRecorderMemory>(nullptr);
+        RefPtr<DrawTarget> canvasTarget =
+            gfx::Factory::CreateRecordingDrawTarget(
+                recorder, referenceDt,
+                gfx::IntRect(gfx::IntPoint(), size.ToUnknownSize()));
         if (!canvasTarget) {
           continue;
         }

But this hits this warning. With that diff, I get there with a mozilla::gfx::SourceSurfaceRecording, which I think it's just not handled. Should it?

Another potential approach would be to maybe use the DrawDependentSurface mechanism that iframes use, instead of CreateSimilarDrawTarget. I haven't tried that yet but it should work as well...

Lee, any suggestion on the preferred approach?

Comment 6

[Emilio Cobos Álvarez [:emilio]]

The good thing is that this is just related to pdf.js / MozPrintCallback and not a more general problem

Comment 7

[Emilio Cobos Álvarez [:emilio]]

Attachment: Bug 2034406 - Support vector-backed similar DrawTargets for skia. r=lsalzman!

This seems to do the trick, but I'm not super-happy about it:

• SourceSurfaceSkia generally assumes non-null mImage, and we are
  adding this mPicture codepath, should I add a SourceSurfaceSkPicture
  or something?

• Similarly, having to null out mCanvas and such after calling
  Snapshot() is a bit hacky, but it works for our purposes.

Comment 8

[Emilio Cobos Álvarez [:emilio]]

Attachment: Bug 2034406 - Preserve vector commands in MozPrintCallback. r=lsalzman!

By teaching DrawTargetRecording to draw a SourceSurfaceRecording
properly.

Comment 9

[Emilio Cobos Álvarez [:emilio]]

Calixte do you know where / how could / should we add some tests for this?

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/4015a65b8bf7
https://hg.mozilla.org/integration/autoland/rev/f3267046f7cb
Preserve vector commands in MozPrintCallback. r=lsalzman

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/aabf9eb2dd63
https://hg.mozilla.org/integration/autoland/rev/f36d4216725e
Revert "Bug 2034406 - Preserve vector commands in MozPrintCallback. r=lsalzman" for causing webdriver failures.

Comment 12

[Serban Stanca [:SerbanS]]

Reverted this because it was causing webdriver failures.

• Revert link
• Push with failures
• Failure Log
• Failure line: TEST-UNEXPECTED-FAIL | /webdriver/tests/classic/print/printcmd.py | test_large_html_document - AssertionError: unknown error (500): [Exception... "Unexpected error" nsresult: "0x8000ffff (NS_ERROR_UNEXPECTED)" location: "JS frame :: resource:///modules/sessionstore/SessionStore.sys.mjs :: ssi_onQuitApplicationGranted :: line 2910" data: no]

Also, please check ths wpt failure.

Comment 13

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/4a0b38bfb466
https://hg.mozilla.org/integration/autoland/rev/64a340fce93d
Preserve vector commands in MozPrintCallback. r=lsalzman

Comment 14

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/96d3f7803ff4
https://hg.mozilla.org/integration/autoland/rev/de4a613bc288
Revert "Bug 2034406 - Preserve vector commands in MozPrintCallback. r=lsalzman" for causing wpt failures

Comment 15

[Sandor Molnar[:smolnar]]

Backed out for causing wpt failures @ mozilla/html/canvas/mozPrintCallback-rect-001-print.html

Backout link: https://hg.mozilla.org/integration/autoland/rev/de4a613bc2881b79bd28b85bce3881907dd914ea

Push with failures

Failures log ->TEST-UNEXPECTED-FAIL | /_mozilla/html/canvas/mozPrintCallback-rect-00x-print.html

Comment 16

[Calixte Denizet (:calixte)]

(In reply to Emilio Cobos Álvarez [:emilio] from comment #9)

Calixte do you know where / how could / should we add some tests for this?

I'd say in test directory for printing stuff.
You can use something like:
https://searchfox.org/firefox-main/source/layout/tools/reftest/reftest.sys.mjs#2240

Once you've a document from getDocument, you can call getPage(1) to get the first page and then call getTextContent in order to make sure that isn't not just an image and assert on the text content.
Jamie did something similar:
https://searchfox.org/firefox-main/source/accessible/tests/browser/pdfOutput/head.js#109-127

Comment 17

[Emilio Cobos Álvarez [:emilio]]

(In reply to Sandor Molnar[:smolnar] from comment #15)

Backed out for causing wpt failures @ mozilla/html/canvas/mozPrintCallback-rect-001-print.html

Was missing a setTransform call.

Comment 18

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/a49a71cc6e3e
https://hg.mozilla.org/integration/autoland/rev/cc4cb03d2653
Preserve vector commands in MozPrintCallback. r=lsalzman

Comment 19

[rperta]

https://hg.mozilla.org/mozilla-central/rev/cc4cb03d2653

Comment 20

[Emilio Cobos Álvarez [:emilio]]

sheriffs: Can we back this out of beta for causing bug 2045512? We can leave it on Nightly.

Comment 21

[Sandor Molnar[:smolnar]]

(In reply to Emilio Cobos Álvarez [:emilio] from comment #20)

sheriffs: Can we back this out of beta for causing bug 2045512? We can leave it on Nightly.

Pascal is current release owner for beta

Comment 22

[Pulsebot]

https://github.com/mozilla-firefox/firefox/commit/1fe0ec5ab479
https://hg.mozilla.org/releases/mozilla-beta/rev/7bc85398f24c

Comment 23

[Ardelean Oana, Desktop Test Engineering [:oardelean]]

Reproducible on a 2026-04-23 Firefox Nightly build on Windows 10.

Verified as fixed on Firefox Nightly 154.0a1 on Windows 10, Ubuntu 22, macOS 13.