wine/gecko
0
0
Fork 0
mirror of https://gitlab.winehq.org/wine/wine-gecko.git synced 2024-09-13 09:24:08 -07:00
Code Issues Packages Projects Releases Wiki Activity

Bug 1089050 part 4. Strengthen the asserts in UpdateExistingSetPropCallStubs now that the stronger asserts should hold. r=jandem

Browse Source
This commit is contained in:
Boris Zbarsky 2014-11-06 23:41:45 -05:00
parent ecb9fb8707
commit cec37cb8ce
1 changed files with 3 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
js/src/jit/BaselineIC.cpp
View File

@ -5766,17 +5766,9 @@ UpdateExistingSetPropCallStubs(ICSetProp_Fallback* fallbackStub,
if (setPropStub->holder() == holder) {
// We want to update the holder shape to match the new one no
// matter what, even if the receiver shape is different.
//
// We would like to assert that either
// setPropStub->holderShape() != holder->lastProperty() or
// setPropStub->shape() != receiverShape, but that assertion can
// fail if there is something in a setter that changes something
// that we guard on in our stub but don't check for before/after
// differences across the set during stub generation. For
// example, a setter mutating the shape of the proto the setter
// lives on would cause us to create an IC stub that never
// matches as protos with the old shape flow into it, but always
// matches post-set, which is where we are now.
MOZ_ASSERT(setPropStub->holderShape() != holder->lastProperty() ||
setPropStub->shape() != receiverShape,
"Why didn't we end up using this stub?");
setPropStub->holderShape() = holder->lastProperty();
// Make sure to update the setter, since a shape change might
// have changed which setter we want to use.