[cfe-commits] [PATCH][MS] - Fix 6 lit tests on windows

Francois Pichet pichet2000 at gmail.com
Wed Sep 8 18:03:08 PDT 2010


I just spent a hour on this. 3 observations:
1- it crashes in fprintf c-index-test.c.
2- it doesn't crash in debug mode. only in release
3- it doesn't seem to crash if executed outside of the python context.

investigation to continue...

crash call stack:

MSVCR90.dll!_invoke_watson+0xf9
MSVCR90.dll!_write+0x59
MSVCR90.dll!_flsbuf+0xd4
MSVCR90.dll!_vcwprintf_s+0x21a2
MSVCR90.dll!_vcwprintf_s+0x2cad
MSVCR90.dll!fprintf+0xf0
c-index-test.exe+0x3275
c-index-test.exe+0x32e0
c-index-test.exe+0x34e5
c-index-test.exe+0x443a




On Wed, Sep 8, 2010 at 5:56 PM, Daniel Dunbar <daniel.dunbar at gmail.com> wrote:
> Have you tried inserting additional code to validate that a < b => !(b
> < a) and so on? Sometimes that is enough to flush out the problem.
>
>  - Daniel
>
>
> On Sep 7, 2010, at 16:57, Douglas Gregor <dgregor at apple.com> wrote:
>
>>
>> On Sep 7, 2010, at 4:52 PM, Francois Pichet wrote:
>>
>>> On Tue, Sep 7, 2010 at 7:29 PM, Daniel Dunbar <daniel at zuster.org> wrote:
>>>> Hi Francois,
>>>>
>>>> I didn't see this email, but I just applied a different fix which uses
>>>> -ffreestanding instead, which avoids introducing a platform specific
>>>> dependency. This resolved all the test failures on VS10, at least for
>>>> me.
>>>
>>> Great this fix 8 tests.
>>> I am still getting some tests failing using VS 2008 and Windows 7 x64
>>> + about 4 popups with c-index-test.exe crashing.
>>>
>>> Failing Tests (8):
>>>   Clang :: Driver/darwin-ld.c
>>>   Clang :: Index/complete-exprs.c
>>>   Clang :: Index/complete-hiding.c
>>>   Clang :: Index/complete-macros.c
>>>   Clang :: Index/complete-natural.m
>>>   Clang :: Index/complete-preprocessor.m
>>>   Clang :: Index/preamble-reparse.c
>>>   Clang :: Index/preamble.c
>>
>> The code-completion failures seem to be due to a bogus partial ordering in the clang::operator< in lib/Sema/CodeCompleteConsumer.cpp, but (from inspection) I don't see any problems there. Very annoying!
>>
>>    - Doug
>




More information about the cfe-commits mailing list