Bug 72382

Seems to be related to an additional '\n' that is being manually added in some cases, at the end of a line. See https://bugs.webkit.org/show_bug.cgi?id=25415#c78. Now working on it...

Created attachment 122263 [details] Patch proposal + Unit test

Comment on attachment 122263 [details] Patch proposal + Unit test View in context: https://bugs.webkit.org/attachment.cgi?id=122263&action=review > Source/WebKit/gtk/tests/testatk.c:796 > +static void testWebkitAtkGetTextAtOffserWithPreformattedText() Obviously, this is a typo. The right name would be testWebkitAtkGetTextAtOffsetWithPreformattedText()

Created attachment 122750 [details] Patch proposal + Unit test New patch, fixing some stupid mistakes in the previous one.

Created attachment 123799 [details] Patch proposal + Unit Test Updating the patch after the refactor done for fixing bug 76783

Comment on attachment 123799 [details] Patch proposal + Unit Test View in context: https://bugs.webkit.org/attachment.cgi?id=123799&action=review > Source/WebCore/accessibility/gtk/WebKitAccessibleInterfaceText.cpp:-105 > - box = box->nextTextBox(); Did you mean to remove this line? I don't see anyting about this in the ChangeLog?

Comment on attachment 123799 [details] Patch proposal + Unit Test While smoke testing some stuff with MiniBrowser and Orca, I found out today that this patch seems to cause some hangs when browsing with the caret, so I'm withdrawing the r? flag for now.

(In reply to comment #7) > (From update of attachment 123799 [details]) > While smoke testing some stuff with MiniBrowser and Orca, I found out today that this patch seems to cause some hangs when browsing with the caret Found an Orca-free way to reliably reproduce the hang: 1. Launch the tester (below) 2. Launch Epiphany 3. Navigate via the caret I get reliable and immediate hangs in Epiphany, along with the following pyatspi lookup error: Traceback (most recent call last): File "/usr/lib/python2.7/site-packages/pyatspi/registry.py", line 193, in eventWrapper return callback(event) File "./caret-moved.py", line 11, in listener print text.getTextAtOffset(event.detail1, pyatspi.TEXT_BOUNDARY_LINE_START) File "/usr/lib/python2.7/site-packages/pyatspi/text.py", line 443, in getTextAtOffset ret = exwrap(Atspi.Text.get_text_at_offset, self.obj, offset, type) File "/usr/lib/python2.7/site-packages/pyatspi/utils.py", line 339, in exwrap raise LookupError LookupError BTW pyatspi LookupErrors are the spawn of <insert your personal devil here>. I'll poke Mike Gorse. -------------------- #!/usr/bin/python import pyatspi def listener(event): try: text = event.source.queryText() except NotImplementedError: return print text.getTextAtOffset(event.detail1, pyatspi.TEXT_BOUNDARY_LINE_START) pyatspi.Registry.registerEventListener(listener, "object:text-caret-moved") pyatspi.Registry.start() --------------------

(In reply to comment #6) > (From update of attachment 123799 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=123799&action=review > > > Source/WebCore/accessibility/gtk/WebKitAccessibleInterfaceText.cpp:-105 > > - box = box->nextTextBox(); > > Did you mean to remove this line? I don't see anyting about this in the ChangeLog? So... I'm going to test having put that line back. Unfortunately, I'm still building (building building) WebKitGtk due to having built Gtk+ and Glib from master. But in the meantime, it sure looks to me that this may be the source of the hang because I don't see anything else updating the value of box which is being looped over.

So finally the build completed. I did some quick testing of WebKit from master with the patch proposed with the accidentally removed line put back. Bug seems to be fixed, hang is definitely gone.

Created attachment 124990 [details] Patch proposal + Unit Test

(In reply to comment #6) > (From update of attachment 123799 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=123799&action=review > > > Source/WebCore/accessibility/gtk/WebKitAccessibleInterfaceText.cpp:-105 > > - box = box->nextTextBox(); > > Did you mean to remove this line? I don't see anyting about this in the ChangeLog? (In reply to comment #10) > So finally the build completed. I did some quick testing of WebKit from master with the patch proposed with the accidentally removed line put back. Bug seems to be fixed, hang is definitely gone. You both were right. Sorry for this mistake, and for making you waste your time... now attached the right one

Committed r106545: <http://trac.webkit.org/changeset/106545>