[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#626021: linux-image-2.6.38-2-686: icebp (opcode 0xf1) no longer causing a SIGTRAP, breaks Wine



On Tue, May 10, 2011 at 07:43:37PM -0500, Austin English wrote:
> On Tue, May 10, 2011 at 19:41, Frederic Weisbecker <fweisbec@gmail.com> wrote:
> > It seems not adapted to 64 bits in fact.
> > The part that complains is in an "#ifdef __i386__", and there
> > is an alternate __x86_64__ section. It seems the section that has been
> > compiled is the i386 even if I'm running amd64, which may be the
> > cause of the issue.
> >
> > So the problem appears to be more in the test implementation of wine.
> 
> Unless you configured wine with --enable-win64 when building, wine is
> being compiled as a 32-bit application.
> 
> Though I'm more worried about the i386 case failing :).

There are few chances that 2.6.32 is concerned by a trap/breakpoints
bug that such a test could detect. The trap/breakpoints bugs have started
in 2.6.33 when we wrote the breakpoint subsystem, which caused some regressions.
If a problem was there before, and it's detected by a wine test, I think it would
have been reported a while ago already.

So I think it's likely a problem in the wine selftest. I just can't find
my way in that code. The address of test_stage is supposed to match some
address given as argument to the test. But test_stage addr has not been
read before that happens, and it's a static var. How could it match anything.
Anyway, I'm definetly clueless there :)

BTW, I wrote a breakpoint/trap regression test a little while ago, in case
that helps finding some regressions. At least it should be able to find the
most obvious one:
	https://tglx.de/~fweisbec/breakpoint_test.c



Reply to: