1731911 - Fix nsinstall parallel build failure

Status: RESOLVED FIXED

(NSS :: Build, defect, P5)

Description

[Giulio Benetti]

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/93.0.4577.82 Safari/537.36

Steps to reproduce:

I've built nss with make -j and rarely it fails.

Actual results:

../../coreconf/config.mk:138: CPU_ARCH is not x86_64, disabling -mavx2
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssck.api ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckepv.h ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckft.h ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckft.h ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckfw.h ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckfw.h ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckfwc.h ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckfwc.h ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckfwt.h ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckfwt.h ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckg.h ../../../dist/public/nss
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckg.h ../../../dist/public/nss
symlink creation race: /home/buildroot/autobuild/run/instance-1/output-1/build/libnss-3.69.1/dist/public/nss/nssckg.h
nsinstall: symlink was attempted in working directory /home/buildroot/autobuild/run/instance-1/output-1/build/libnss-3.69.1/nss/lib/ckfw from ../../../nss/lib/ckfw/nssckg.h to /home/buildroot/autobuild/run/instance-1/output-1/build/libnss-3.69.1/dist/public/nss/nssckg.h.
: File exists
../../coreconf/nsinstall/Linux2.6_ppc_powerpc64le-linux-gcc.br_real_glibc_PTH_64_DBG.OBJ/nsinstall -R -m 444 nssckmdt.h ../../../dist/public/nss
make[4]: *** [../../../dist/public/nss/nssckg.h] Aborted

Comment 1

[Giulio Benetti]

Attachment: 0001-Bug-1731911-Fix-nsinstall-parallel-failure.patch

I know this is not the best fix but it should work. Unfortunately failure happens very rarely and I can't reproduce, so I ask to commit to see what happens. Same thing I will ask in Buildroot, so we will have their autobuilders testing it. Any other suggestions is welcome.

Best regards

Comment 2

[Giulio Benetti]

It's almost 3 days that builds continuosly and no error shows up.

Comment 3

[Benjamin Beurdouche [:beurdouche]]

@Giulio, do you still see this problem ?

Comment 4

[Giulio Benetti]

@Benjamin, yes, also with nss 3.71. Same thing.
I’ve let it building it continuously for 10 days with this patch
and I’ve had no problems. It’s also true I couldn’t reproduce
the error honestly.
Here is another failure:
http://autobuild.buildroot.net/results/a5f/a5f8a126a42ab784c3b5ab375a8cd7cbf205d745/build-end.log

Buildroot maintainers still didn’t apply this patch.

Comment 5

[Benjamin Beurdouche [:beurdouche]]

@Bob, I am not an expert but this looks like a race condition in the makefiles to me.
Any chance you have someone at RH that can have a look at this?

Comment 6

[Robert Relyea]

I think the patch is fine nsinstall needs to be built before you kick off all the rest of the stuff, so it's OK to set it to noparrellel.
We (Red Hat) don't do parrallel builds right now, so we haven't ran into the issue.

Comment 7

[Benjamin Beurdouche [:beurdouche]]

Attachment: Bug 1731911 - Fix nsinstall parallel failure

Comment 8

[Benjamin Beurdouche [:beurdouche]]

https://hg.mozilla.org/projects/nss/rev/f7b146b603f76c2754877abd65542b0b571e4b9d
Thanks Giulio!