290182 - firefox nightly trunk builds crashes at start with the localized resouces [@ MSVCRT.DLL - nsTextBoxFrame::UpdateAttributes ]

Status: VERIFIED FIXED

(Core :: XUL, defect)

Description

[Shaohua Wen]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; zh-CN; rv:1.7.7) Gecko/20050403 Firefox/1.0.3
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.0; zh-CN; rv:1.7.7) Gecko/20050403 Firefox/1.0.3

I downloaded firefox nightly trunk builds (4.13 builds,and 4.09 builds),and I
tried to translate the resouces (.dtd&properties) to Chinese (Simplified),
during the testing,the localized builds crash at start....
 
I tried to find out the problem,and found that if in the localized resource
files,there is an accesskey or accessCommad,the application will crash when
open the UI contains it..

Here I attach the browser.dtd with only one entry translated and you can 
just put it into the en-US.jar and replace the original and the application will
crash immediatly.

In this case Firefox is not localizable any more...


Reproducible: Always

Comment 1

[Shaohua Wen]

Attachment: browser.dtd with only one entry translated to Chinese

Comment 2

[Benjamin Smedberg]

1) I'm confused. The attachment looks like you only translated toolsMenu.label.
Is this correct? This is not an accesskey or commandkey.

2) Can you get a talkback report number from your crash?

Comment 3

[Shaohua Wen]

in case toolsMenu.label is translated and there is a toolsMenu.accesskey it will
crash.
for other entries,it will not crash if there is only xxx.label.

I cannot send the talkback because the talkback agent cannot send the report
through 
our http proxy (i checked the settings are correct,but it just not working).

Comment 4

[Shaohua Wen]

OK it works this time... :)
the talk back number is:
TB5062930H

Comment 5

[Benjamin Smedberg]

In a debug build (using MSVC.NET2003), I get a MSVC runtime assertion at the
isspace() call at
http://bonsai.mozilla.org/cvsblame.cgi?file=/mozilla/layout/xul/base/src/nsTextBoxFrame.cpp&mark=710&rev=#700
Because we're passing it a value > 256. This is probably a unicode mismatch of
some sort. However, if I continue past the assertion my build runs fine, and so
does a release build.

Comment 6

[Shaohua Wen]

So when can we get an localizable nightly builds?
Thanks!

Comment 7

[jayfromtaiwan]

Same crash (supposedly) with a VC8 beta2-built firefox build. --> msvcr80.dll

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b2) Gecko/20050520
Firefox/1.0+ (JTw)

Comment 8

[jayfromtaiwan]

(In reply to comment #7)
> Same crash (supposedly) with a VC8 beta2-built firefox build. --> msvcr80.dll
> 
> Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b2) Gecko/20050520
> Firefox/1.0+ (JTw)

Tried a custom build ... confirmed that backing out the patch for Bug 260563
eliminates the crash

Comment 9

[timeless]

http://msdn.microsoft.com/library/default.asp?url=/library/en-us/vclib/html/_crt_isspace.2c_.iswspace.asp

Return Value

isspace returns a non-zero value if c is a white-space character (0x09 – 0x0D or
0x20). The result of the test condition for the isspace function depends on the
LC_CTYPE category setting of the current locale; see setlocale for more information.

iswspace returns a non-zero value if c is a wide character that corresponds to a
standard white-space character.

When used with a debug CRT library, isspace will display a CRT assert if passed
a parameter that isn't EOF or in the range of 0 through 0xFF. When used with a
debug CRT library, isspace will use the parameter as an index into an array,
with undefined results if the parameter isn't EOF or in the range of 0 through 0xFF.

undefined results include crashing.

NS_IS_SPACE should do the right thing.

Comment 10

[timeless]

Attachment: be nice to vcrt

Comment 11

[Daniel Veditz [:dveditz]]

Comment on attachment 187106 [details] [diff] [review]
be nice to vcrt

r/sr=dveditz

Comment 12

[Daniel Veditz [:dveditz]]

Comment on attachment 187106 [details] [diff] [review]
be nice to vcrt

a=dveditz for trunk

Comment 13

[timeless]

Comment on attachment 187106 [details] [diff] [review]
be nice to vcrt

mozilla/layout/xul/base/src/nsTextBoxFrame.cpp	1.78

Comment 14

[timeless]

filed bug 298612 as followup

reporter: please verify (or reopen if this isn't really fixed)

Comment 15

[Shaohua Wen]

tested with the firefox nightly builds (date 7.02).
now it works!
Thanks!