Bug 238084 - r348115 causes reproducable compiler crash in buildworld or buildkernel
Summary: r348115 causes reproducable compiler crash in buildworld or buildkernel
Status: Closed FIXED
Alias: None
Product: Base System
Classification: Unclassified
Component: kern (show other bugs)
Version: CURRENT
Hardware: amd64 Any
: --- Affects Only Me
Assignee: freebsd-bugs (Nobody)
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2019-05-23 21:40 UTC by rz-rpi03
Modified: 2019-05-24 19:14 UTC (History)
2 users (show)

See Also:

Attachments
Compiler call from make buildworld which produces a c++.core (4.61 KB, text/plain)
2019-05-23 21:40 UTC, rz-rpi03 		no flags Details
gziped source file (350.03 KB, application/gzip)
2019-05-24 07:06 UTC, rz-rpi03 		no flags Details
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description rz-rpi03 2019-05-23 21:40:12 UTC 
Created attachment 204578 [details]
Compiler call from make buildworld which produces a c++.core

After upgrading from r348059 to r348117 make buildworld fails with a compiler core dump.
Rebooting with the r348059 kernel and using the r348117 world fixes this.

While bisecting I learned that make buildkernel fails as well if the bug is
there.

Finally the change in base r348115 (review D20356) was the cause.

With a r348114 kernel, the attached ARMTargetParser-5d7a90.sh generates a
ARMTargetParser-5d7a90.o file.
With a r348115 kernel a c++.core.

The corresponding ARMTargetParser-5d7a90.cpp is to big for a attachment
because it is 2.4 MB in size.

Regards, Ralf
Comment 1 rz-rpi03 2019-05-24 07:06:26 UTC 
Created attachment 204587 [details]
gziped source file

The corresponding gziped source file is small enough to be added as attachment.
Comment 2 Shawn Webb 2019-05-24 15:36:42 UTC 
HardenedBSD is also affected.
Comment 3 Mark Johnston freebsd_committer freebsd_triage 2019-05-24 15:38:26 UTC 
Have you tried updating to r348128?
Comment 4 rz-rpi03 2019-05-24 19:14:01 UTC 
(In reply to Mark Johnston from comment #3)

No, the crash occured first during a buildworld for a newer release than r348117. So I stopped going forward.

Just tried the r348128 kernel - build with r348114 - and the compiler crash
with the attached files is gone. At the moment a r348248 buildworld is past
the point where the crash happend. I am closing this bug report.

Thank you.