PR/57241 CVS commit: src

["Taylor R Campbell" <riastradh%netbsd.org@localhost>]

The following reply was made to PR toolchain/57241; it has been noted by GNATS.

From: "Taylor R Campbell" <riastradh%netbsd.org@localhost>
To: gnats-bugs%gnats.NetBSD.org@localhost
Cc: 
Subject: PR/57241 CVS commit: src
Date: Wed, 17 Dec 2025 01:32:55 +0000

 Module Name:	src
 Committed By:	riastradh
 Date:		Wed Dec 17 01:32:55 UTC 2025
 
 Modified Files:
 	src/libexec/ld.elf_so: Makefile
 	src/share/mk: bsd.prog.mk
 
 Log Message:
 bsd.prog.mk: Fix parallel builds of debug data.
 
 Previously, we had one rule to generate foo, and another rule to
 derive foo.debug from it with objcopy -- and then rewrite foo _in
 place_ to strip the debug data with objcopy.
 
 This is wrong -- one rule should never overwrite another rule's
 target; this violates the contract with make(1), and can lead it to
 run rules in parallel on files that are changing, which in turn can
 lead the rules to behave mysteriously or crash.
 
 It's still not clear why in our autobuilds, the only cases where we
 saw these crashes were in mips64 builds of programs that use
 compat/exec.mk (mainly external/bsd/ipf/bin/ipftest, but also
 usr.sbin/crash and usr.bin/systat).  But the bug this change fixes is
 well-understood, and now someone has observed it in a non-mips build,
 so I'm throwing in the towel on figuring out what makes these
 particular programs more likely to trigger the problem.
 
 PR toolchain/57241: mips64el--netbsd-install core dumps randomly
 
 
 To generate a diff of this commit:
 cvs rdiff -u -r1.151 -r1.152 src/libexec/ld.elf_so/Makefile
 cvs rdiff -u -r1.359 -r1.360 src/share/mk/bsd.prog.mk
 
 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.