summaryrefslogtreecommitdiffstats
path: root/base/md5.cc
diff options
context:
space:
mode:
authordcheng@chromium.org <dcheng@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98>2013-02-23 01:56:14 +0000
committerdcheng@chromium.org <dcheng@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98>2013-02-23 01:56:14 +0000
commitb102bbd622753faeb7440d59174fdedfb70c54da (patch)
tree12fbc8b3517a7a6079526d0c01849a532f564147 /base/md5.cc
parentbf91227ca2796661a6b17176dc50f4db002a7b49 (diff)
downloadchromium_src-b102bbd622753faeb7440d59174fdedfb70c54da.zip
chromium_src-b102bbd622753faeb7440d59174fdedfb70c54da.tar.gz
chromium_src-b102bbd622753faeb7440d59174fdedfb70c54da.tar.bz2
Revert 184179
This is trigger a latent issue in views which results in a memory leak each time a Widget is destroyed. > Update scoped_ptr<T>::reset() to more closely match std::unique_ptr<T>. > > Remove the no-op behavior of a scoped_ptr "self-reset", and update the > reset() method to detect problematic dependencies on the sequence of > events. Eventually, reset() will be updated to exactly match the > implementation of std::unique_ptr<T> by setting data_.ptr to the new > value before deleting the old value. > > However, this will expose latent bugs where a destructor invoked > transitively by reset() attempts to dereference the same scoped_ptr. > Relying on the value of get() in this instance will dispatch calls to > the incorrect object. As a temporary measure to detect this class of > bugs, set data_.ptr to NULL during deletion so that it results in a > crash. > > BUG=162971,176091 > > Review URL: https://codereview.chromium.org/12223113 TBR=dcheng@chromium.org Review URL: https://codereview.chromium.org/12317084 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@184255 0039d316-1c4b-4281-b951-d872f2087c98
Diffstat (limited to 'base/md5.cc')
0 files changed, 0 insertions, 0 deletions