Bug 124387 - [ASAN] Ability to disable all ASSERTs except for ASSERT_WITH_SECURITY_IMPLICATION
Summary: [ASAN] Ability to disable all ASSERTs except for ASSERT_WITH_SECURITY_IMPLICA...
Status: RESOLVED DUPLICATE of bug 125329
Alias: None
Product: WebKit
Classification: Unclassified
Component: Tools / Tests (show other bugs)
Version: 528+ (Nightly build)
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: David Farler
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-11-14 16:28 PST by David Farler
Modified: 2014-01-30 11:39 PST (History)
5 users (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description David Farler 2013-11-14 16:28:38 PST
Since Debug builds are going to be catching assertions anyway and ASAN builds are more geared toward automation, we should at least set ASSERT_DISABLED=1 for the builds in the asan.xcconfig.
Comment 1 David Farler 2013-11-14 16:33:14 PST
It also looks like there is already a convenient ifdef for ADDRESS_SANITIZER which we can use for ASSERT_WITH_SECURITY_IMPLICATION.
Comment 2 David Farler 2013-11-14 16:55:39 PST
Hm, looks like ASSERT_DISABLED=1 and ADDRESS_SANITIZER -> ASSERT_WITH_SECURITY_IMPLICATION is an invalid combination because isInAtomicStringTable is wrapped with #if !ASSERT_DISABLED so the build fails with a missing implementation.
Comment 3 David Farler 2013-11-14 17:10:23 PST
Ah, I was wrong. ASSERTIONS_DISABLED_DEFAULT =1 is the top level. However, this gets reset to 0 ifndef NDEBUG (i.e. for Debug builds).

There are several categories underneath that are controlled by the top level ASSERTIONS_DISABLED_DEFAULT, so I'll see how I can do with setting those. It looks like those are the only places where ASSERTIONS_DISABLED_DEFAULT is accessed.
Comment 4 David Kilzer (:ddkilzer) 2013-11-15 14:23:10 PST
Is the goal to build with ASSERT_DISABLED=1, but with ASSERT_WITH_SECURITY_IMPLICATION enabled?  I think that would be a very useful combination.
Comment 5 David Farler 2013-11-15 15:08:22 PST
I'm finding out the hard way that no one has recently, if ever, broken the assumption that Debug -> Asserts. There are some mismatches with #ifndef NDEBUG and #if !ASSERT_ that cause build/link failures.
Comment 6 David Kilzer (:ddkilzer) 2014-01-25 20:53:43 PST
I think this was fixed by another commit.

Maybe commit r160462 for Bug 125329.
Comment 7 David Farler 2014-01-30 11:39:04 PST
Yep. Closing.

*** This bug has been marked as a duplicate of bug 125329 ***