[llvm-bugs] [Bug 49185] opt -O3 crashes with Assertion `NewL->getParentLoop() == ParentL && "All of the new loops must be siblings of the current loop!"' failed. with new PM

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=49185

Mikael Holmén <mikael.holmen at ericsson.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED

--- Comment #6 from Mikael Holmén <mikael.holmen at ericsson.com> ---
(In reply to Arthur Eubanks from comment #5)
> There's no way to get a list of passes for -O3 under the new PM.
> 

Almost every time I debug an opt problem I start with "opt -O<N> <something
more>" and I want to reach an "opt <singlepass>" reproducer to get an as small
and stable reproducer as possible that isolates the real problem. Normally I
use -debug-pass=Arguments along with reducing the input program, and in most
cases I end up with what I want.

How do we achieve this with the new PM then? I'm sure people already working
with the new PM want to do the same thing, so how do you do?

I already sent an email to llvm-dev with this question but unfortunately got
zero replies so far:
https://lists.llvm.org/pipermail/llvm-dev/2021-February/148551.html

> I can't repro the crash locally at head. Perhaps it's fixed? Or maybe I'm
> not specifying the right backend target since it's not specified in the .ll
> file.

Yes it was fixed in f70cdc5b5c.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20210224/b846c653/attachment-0001.html>