Bug 70725 – [Qt][WK2] Crash when running pixel tests

• Component: WebKit2
• Status: RESOLVED FIXED

[Balazs Kelemen]

Linux, 64-bit, up-to-date qt5.

#0  0x00007f14d526f509 in memcpy () from /lib64/libc.so.6
#1  0x00007f14d62b5849 in MD5Update (ctx=0x764188, buf=0xc3cff0 "", len=2304) at tools/../../3rdparty/md5/md5.cpp:95
#2  0x00007f14d62b8b27 in QCryptographicHash::addData (this=0x7fff5f3ec840, data=
    0x9cc8f0 "\377\377\377\377...
#3  0x00000000004096ea in WTR::TestInvocation::dumpPixelsAndCompareWithExpected(OpaqueWKImage const*, OpaqueWKArray const*) ()
#4  0x000000000040c140 in WTR::TestInvocation::didReceiveMessageFromInjectedBundle(OpaqueWKString const*, void const*) ()
#5  0x00007f14d986d71e in WebKit::WebContextInjectedBundleClient::didReceiveMessageFromInjectedBundle(WebKit::WebContext*, WTF::String const&, WebKit::APIObject*) () from /home/balazs/work/master_clean/wk2BUild/Release/bin/../lib/libQtWebKit.so.4

[Balazs Kelemen]

Oops, my fault in  http://trac.webkit.org/changeset/94524:
hash.addData(reinterpret_cast<const char*>(image.constScanLine(row)), image.width() * image.bytesPerLine());

|width * bytesperline| is a bit too much ...

[Balazs Kelemen]

Created attachment 112207 [details]
Patch

[Chang Shu]

Comment on attachment 112207 [details]
Patch

Just curious how this worked before. In addition, bytesPerLine sounds like it has taken width into consideration.

[Chang Shu]

Comment on attachment 112207 [details]
Patch

sorry, i thought the code was added but it was removed. r=me.

[Balazs Kelemen]

Comment on attachment 112207 [details]
Patch

Clearing flags on attachment: 112207

Committed r98252: <http://trac.webkit.org/changeset/98252>

[Balazs Kelemen]

All reviewed patches have been landed.  Closing bug.

[Balazs Kelemen]

(In reply to comment #4)
> (From update of attachment 112207 [details])
> sorry, i thought the code was added but it was removed. r=me.

Sure, the value we passed was way too high, but sometimes it was working well with memory garbage :)
Thanks for the review.