1429812 - Investigate MSVC's Return Flow Guard

Status: RESOLVED WONTFIX

(Core :: Security, enhancement)

Description

[Tom Ritter [:tjr] (OOTO until 4/30 at least)]

/guard:rf

Literally brand new in 15.5.3 so no one knows what it does or how it works. MSFT will release more information eventually, I just don't want to forget about it.

Comment 1

[Tom Ritter [:tjr] (OOTO until 4/30 at least)]

Try run:  https://treeherder.mozilla.org/#/jobs?repo=try&revision=cd0d24d4447b42c9bf013d3b65940dfad746246f

Comment 2

[Tom Ritter [:tjr] (OOTO until 4/30 at least)]

Perf results are in: https://treeherder.mozilla.org/perf.html#/compare?originalProject=mozilla-central&newProject=try&newRevision=cd0d24d4447b42c9bf013d3b65940dfad746246f&framework=1&selectedTimeRange=172800

It's pretty sizable regressions all across the board. Except for stylebench for some weird reason.

Comment 3

[Tom Ritter [:tjr] (OOTO until 4/30 at least)]

Reminder to self; this isn't going to be useful unless platform libs are compiled with it....

Comment 4

[(Away)]

Do you expect clang to implement this at some point? If it's strictly MSVC-only then it's pretty much wontfix at this point.

Comment 5

[Tom Ritter [:tjr] (OOTO until 4/30 at least)]

(In reply to David Major [:dmajor] from comment #4)

Do you expect clang to implement this at some point? If it's strictly MSVC-only then it's pretty much wontfix at this point.

Hard to guess. For reference, CFG : CFI :: RFG :: Bug 1465859 (which they removed.)

Comment 6

[(Away)]

We should probably just close this out rather than leave it pending an unknown situation. I'm sure that if clang does end up doing something similar, we'll be quick to open a bug (or revisit this one) as soon as we hear about it.