Benjamin Smedberg Assignee
	Description • 15 years ago

Print preview is broken in electrolysis (all remoting off). Not sure which patch to blame yet, but it's likely to be one of the nsFrameLoader ones, so this doesn't block landing IPC plugins in mozilla-central. It is keeping the e10s windows tinerbox orange, though (we have print-preview tests)!

Possible candidates for causing the regression: bug 522352 (although most of that landed in mozilla-central as well), bug 505847, unlikely but possibly bug 523224 or bug 516727

	Benjamin Smedberg Assignee
	Comment 1 • 15 years ago

This is the diff from mozilla-central to e10s in widget, content, and docshell. Does anything stand out?

	Boris Zbarsky [:bzbarsky]
	Comment 2 • 15 years ago

My gut feeling is bug 523224.  Not sure why that would do it, though, after reading through that patch again.

Does print preview go through the !IsDynamic() codepath in nsSubDocumentFrame:ShowViewer?

	Olli Pettay [:smaug][bugs@pettay.fi]
	Comment 3 • 15 years ago

So this problem happens only on Windows, and only when running
mochitest (or chrome/browser-chrome).
When I run e10s normally, print preview does work.

	Benjamin Smedberg Assignee
	Comment 4 • 15 years ago

The changes required in nsDeviceContextSpecWin.cpp from bug 522414 look pretty suspicious, actually: http://hg.mozilla.org/projects/electrolysis/rev/478360ac91c4#l6.1

Not that these particular changes are wrong, but if that file is now being compiled with -DUNICODE, it's relatively likely that there's some sort of bad cast going on there. I'll do a quick audit of that file.

	Benjamin Smedberg Assignee
	Comment 5 • 15 years ago

The particular bug here is http://hg.mozilla.org/projects/electrolysis/file/tip/widget/src/windows/nsDeviceContextSpecWin.cpp#l215 where we're casting a LPTSTR to a char* and then converting it back to UTF16.

	Benjamin Smedberg Assignee
	Comment 7 • 15 years ago

Blocks landing OOPP in m-c because this is not OOP-tab-specific.

	Jim Mathies [:jimm]
	Comment 8 • 15 years ago

Comment on attachment 411958 [details] [diff] [review]
Fix UNICODE stuff by using explicit WSTR/WCHAR and avoiding bad casts, rev. 1

That old code looked ancient.

	Ben Turner (not reading bugmail, use the needinfo flag!)
	Comment 9 • 15 years ago

Landed on electrolysis, changeset a5cedf5e76de.