Revert 203172 "NaCl: Update revision in DEPS, r11406 -> r11424"

Revert 203172 "NaCl: Update revision in DEPS, r11406 -> r11424"

This regressed Linux sizes: "PERF_REGRESS: nacl_helper-data/data (1.98%), nacl_helper-data/data (1.98%)"

> NaCl: Update revision in DEPS, r11406 -> r11424
> 
> This pulls in the following Native Client changes:
> 
> r11407: (mseaborn) PNaCl: Update LLVM revision in pnacl/DEPS
> r11408: (mseaborn) PNaCl: Disable ABI checker for GCC torture tests that use inline assembly
> r11409: (bradnelson) Don't use the new buildbot supplied BOTO credentials.
> r11410: (bsy) pread/pwrite implementation. NaClHostDesc level, plus NaClDesc plumbing (2nd try)
> r11411: (mseaborn) PNaCl: Factor out use of {skip_,}nonstable_bitcode flags into method
> r11412: (mcgrathr) toolchain_build: Update gcc revision
> r11413: (sehr) Rename expected exit status from untrusted_sigill to
> r11414: (mseaborn) PNaCl: Update LLVM revision in pnacl/DEPS
> r11415: (shcherbina) Validator exhaustive tests: reorganize superinstruction checking a little bit.
> r11416: (shcherbina) Validator exhaustive test: move regex superinstruction checkers to spec.py.
> r11417: (khim) Fix PRESUBMIT.py to make it compatible with GIT
> r11418: (mseaborn) Update PNaCl toolchain to get r11407: Enable the ReplacePtrsWithInts pass
> r11419: (dschuff) Update binutils rev to bring in size shrink
> r11420: (dschuff) PNaCl: Update LLVM revision in pnacl/DEPS
> r11421: (mseaborn) PNaCl: Update Spec2k performance expectations after enabling ReplacePtrsWithInts
> r11422: (bradnelson) Cleanup code coverage, switch mac code coverage to 64-bit.
> r11423: (bradnelson) Dropping coverage threshold to 49%.
> r11424: (bradnelson) Attempting to fix windows code coverage.
> 
> BUG=none
> TEST=nacl_integration
> 
> Review URL: https://codereview.chromium.org/15914014

TBR=mseaborn@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=203177

[Ilya Sherman, 2013-05-30 18:01:58]

Committed patchset #1 manually as r203177.

[Mark Seaborn, 2013-05-30 18:25:41]

> Revert 203172 "NaCl: Update revision in DEPS, r11406 -> r11424"
>
> This regressed Linux sizes: "PERF_REGRESS: nacl_helper-data/data (1.98%),
nacl_helper-data/data (1.98%)"

I thought the policy was to update perf expectations when this happens, not to
revert changes that happen to pass an arbitrary threshold?  What should I do
here?  Just recommit the change?

[Ilya Sherman, 2013-05-30 18:30:03]

On 2013/05/30 18:25:41, Mark Seaborn wrote:
> > Revert 203172 "NaCl: Update revision in DEPS, r11406 -> r11424"
> >
> > This regressed Linux sizes: "PERF_REGRESS: nacl_helper-data/data (1.98%),
> nacl_helper-data/data (1.98%)"
> 
> I thought the policy was to update perf expectations when this happens, not to
> revert changes that happen to pass an arbitrary threshold?  What should I do
> here?  Just recommit the change?

The policy is to give due consideration to whether the perf regression is
acceptable; and if so, to update the perf expectations.  If we just always
blindly update perf expectations, then we might as well remove the perf checking
altogether.  Please work with the perf sheriffs, pasko and rtenneti, to
understand the regression and reland this change safely.