checking for post-dominators

Thanks

The test passes on mac without xargs. Could you verify windows?

removed xargs

PID is one of the reason there's '*' in compiler-rt tests: runtime information is often dumped in <basename>-<pid> files. These files are then fed to analysis tools by using '<basename>-*' shell glob.

In D30371#686718, @mehdi_amini wrote:

Oh right!

I don't know what the intent was but this is the name of the binary, so it is likely that it use to expand to itself only?

I am confused you say glob expansion happens before other substitution. Shouldn't it be in the opposite order if you want %[[%T/*]] to work?

Full --help output with this change: https://gist.github.com/anonymous/9b20cc6e13b53cf923bf286f0a39972b

Will reupload to add llvm-commits correctly

using target triple as intended.

Actually, hold on a second, I'll replace the triple in darwin file, and that should take care of everything.

Kostya, please take a look.

split trace-pc-guard test into two.

Grhhh. Fails on mac :(

XFAIL on tsan.

replaced XFAIL with REQUIRES.

I've added the test. PTAL.

added .ll test

added at least.

reworded to say may be called more than once.

I think that keeping more general "may be called multiple times" is still correct and does give us room to maneuver.

Related test and documentation update:

update

Updated ctor registration.

LG then.

If this script working under python3 is your goal, then I suggest you make the script work first, rather then try to fix lint issues separately. I'd rather we do python3-ification in one change if possible.

Could you please clarify if you test it with python3 and if it works?

Does this script actually run with python3 after these changes, or are these lint-driven errors? I'm not an expert in python3 (only used python2), but afaik it doesn't have % operator (you are supposed to use "".format()).

I think we can clean this up a bit.

Do you need help landing it?

Another wild thought: is there any template magic that we might invoke to automate creation of function aliases?

I suggest to make WIN* macros no-op on all platforms that do not need it.

I don't like that we introduce macros making function definition ugly and opaque. All this to implement one platform hacks. I would prefer to have current version cleaned up to eliminate ifdefs. Something like:

moved macro to FuzzerDefs.

adding rtti options (for sanitizer flags).

Is this what you had in mind?

This is all about code duplication, right? When you use static libraries - linking gives you an error. When you use dynamic libraries the code is still duplicated, right? Each .so file will still have its own copy of common functions and data. But loader will just pick the first version.

I worry that our set of static and shared runtimes will be different. If we do this split, we should do it for dynamic libraries too. I don't know if we have static initializers, but we might end up running them twice.