media and unit tests failing on XP bot since r780-r790, updated for fcache shifting fix

_From [rnk@google.com](https://code.google.com/u/rnk@google.com/) on March 21, 2012 12:47:46_

XP bot before: http://build.chromium.org/p/chromium.fyi/builders/Windows%20Tests%20%28DrMemory%20XP%29/builds/1027 XP bot after: http://build.chromium.org/p/chromium.fyi/builders/Windows%20Tests%20%28DrMemory%20XP%29/builds/1028 In http://crrev.com/125634 I updated DrMemory in Chromium from r780 to r790 to pull in the fcache shifting bug fix. There were no other changes to media libraries, so that means my update is the most likely culprit. In that drmemory rev range, we updated DR from r1276 to r1286 .

This lead to a bunch of new test failures:
media:
PipelineIntegrationTest.BasicPlayback (crash)
YUVConvertTest.RGB24ToYUV (fail)
YUVConvertTest.SideBySideRGB (fail)

base:
new GDI usage error (XP only it seems)

unit:
JsonSchemaTest.TestNumber (fail)
ProfileSyncServiceTypedUrlTest.HasNativeWithBlankEmptySync (crash)
JsonSchemaTest.TestIntegerBounds (fail)
ProfileSyncServicePasswordTest.FailModelAssociation (crash)

DrMemory changes: r790 Update DynamoRIO to r1286 to pick up the fix for DR issue #696/DrMem issue #784. r789 issue #98, add NtAlpc\* syscall handling, cont'd + continuing progress on NtAlpc syscalls r788 update docs for feature list and light mode r787 fixes issue #805 + automatically disable -callstack_use_top_fp for MinGW apps r786 issue #816 logs and assert (DR r1281 in r783 fixed the issue) r785 fixes issue #813 + iterate over native mallocs too on heap destruction to catch LFH blocks r784 fixes issue #812 + un-replace instead of un-wrap routines that were replaced r783 fixes issue #816 update to DR r1281 r782 fixes issue #775 - add check to handle unaligned-one-byte access in pattern mode r781 fixes issue #643 - add free/delete/delete[] collapse checks r780 set default -light option as -no_check_uninitialized and -no_count_leaks

DR changes: https://code.google.com/p/dynamorio/source/detail?r=1286 Disable fragment shifting when clients are present https://code.google.com/p/dynamorio/source/detail?r=1285 Fix building with DEBUG and without DEADLOCK_AVOIDANCE https://code.google.com/p/dynamorio/source/detail?r=1284 remove unnecessary get_peb calls https://code.google.com/p/dynamorio/source/detail?r=1283 Check the private PEB heap before calling NtQueryInformationProcess https://code.google.com/p/dynamorio/source/detail?r=1282 fixes issue #695 - uninit read error https://code.google.com/p/dynamorio/source/detail?r=1281 fix x64 VS2008 build warning from r1272 https://code.google.com/p/dynamorio/source/detail?r=1280 fixes DrMem issue #816 + update user_data and flags when re-enabling previously disabled wrap https://code.google.com/p/dynamorio/source/detail?r=1279 + prevent confusing errors when building Debug config https://code.google.com/p/dynamorio/source/detail?r=1278 fix misspelling in docs https://code.google.com/p/dynamorio/source/detail?r=1277 Propagate OutputDebugString note from intro.dox to here

Most of these look pretty innocuous. Pulling out the ones that are most likely: r789 issue #98, add NtAlpc\* syscall handling, cont'd + continuing progress on NtAlpc syscalls r785 fixes issue #813 + iterate over native mallocs too on heap destruction to catch LFH blocks r784 fixes issue #812 + un-replace instead of un-wrap routines that were replaced r781 fixes issue #643 - add free/delete/delete[] collapse checks https://code.google.com/p/dynamorio/source/detail?r=1286 Disable fragment shifting when clients are present https://code.google.com/p/dynamorio/source/detail?r=1280 fixes DrMem issue #816 + update user_data and flags when re-enabling previously disabled wrap

Primarily I suspect some fcache shifting handling bug or some drwrap bug.

I'm going to exclude the failing tests for now. I don't have an XP VM readily available, so it will take me awhile to get to this, if anyone wants to take a look.

_Original issue: http://code.google.com/p/drmemory/issues/detail?id=842_