diff options
author | Vlad Tsyrklevich <vtsyrklevich@chromium.org> | 2018-01-04 16:39:45 -0800 |
---|---|---|
committer | Chromium commit bot <commit-bot@chromium.org> | 2018-01-08 15:06:51 +0000 |
commit | dc36f99188feec82454f62d5f489f358dc5fd9c1 (patch) | |
tree | 9a505d3183a445435a84f7a367c34fdfaa9cc26f /fxjs | |
parent | ec24b2e338de2a6211723f19f54386c950ac5010 (diff) | |
download | pdfium-dc36f99188feec82454f62d5f489f358dc5fd9c1.tar.xz |
Reland "[CFI] Enable type generalization for LCMS"
This relands commit 7829b7048d5cf13a4f1a806fed57f2440d1b5dec.
This uses a new version of clang with a fix for the LSan issue that
caused the previous revert.
Original change's description:
> [CFI] Enable type generalization for LCMS
>
> Control Flow Integrity [1] indirect call checking verifies that function
> pointers only call valid functions with a matching type signature. This
> condition can be too strict, a common form of 'abstraction' relies on
> function pointers being cast to generalize argument pointer types to
> void*.
>
> In LCMS these failures occur because of casts of function pointers called
> as both _cmsInterpFn16 (cmsInterpFunction.Lerp16) and _cmsOPTeval16Fn (as
> an argument to _cmsPipelineSetOptimizationParameters) types making it
> difficult to refactor out easily. Instead, enabling the type
> generalization build config weakens the type checking performed for
> CFI-icall to accomodate this common type of casts.
>
> [1] https://www.chromium.org/developers/testing/control-flow-integrity
>
> Bug: 785442
> Change-Id: Ib42fb1b4e152d5042b170698c2707ebb9e0cc1ee
> Reviewed-on: https://pdfium-review.googlesource.com/19250
> Commit-Queue: dsinclair <dsinclair@chromium.org>
> Reviewed-by: dsinclair <dsinclair@chromium.org>
Bug: 785442
Change-Id: Ice696e4f496fa380f8af053577bde7e46c049590
Reviewed-on: https://pdfium-review.googlesource.com/22265
Reviewed-by: dsinclair <dsinclair@chromium.org>
Commit-Queue: dsinclair <dsinclair@chromium.org>
Diffstat (limited to 'fxjs')
0 files changed, 0 insertions, 0 deletions