Created attachment 178082 [details] v0 ports r20327 added an unsafe optimization for caching ARCH value. Rather than run uname(1) let's use MACHINE_ARCH instead as both query sysctl hw.machine_arch, at least in bmake.
I guess I'm confused. I don't see the annotation for r20327. Is this possibly a typo?
Just a bug in our bugzilla configuration. Here again: ports r20327
When cross building, poudriere sets ARCH=${MACHINE_ARCH} in make.conf Does this have an impact with your patch? I guess not yet?
Variables in make.conf don't clobber values in vender Makefile(s) unlike .MAKEFLAGS or those passed on command-line. The exp-run here is to catch unknown dependencies on such a behavior. |make index| maybe a bit faster but not beyond ~ 1% if not at all lost in the noise. > When cross building, poudriere sets ARCH=${MACHINE_ARCH} in make.conf It's going to be the default for everyone.
Comment on attachment 178082 [details] v0 Adding bdrewery for review to avoid rubberstamping after exp-run. If you're curious he did promise to fix the issue... a year ago. https://www.mail-archive.com/freebsd-ports@freebsd.org/msg66955.html
Take for exp-run. Review is still welcome.
Exp-run is fine on 11.0 amd64 and 11.0 i386
(In reply to Antoine Brodin from comment #7) > Exp-run is fine on 11.0 amd64 and 11.0 i386 Looking at https://pkg-status.freebsd.org I'm not convinced queueing mere 1164 packages is enough to qualify for an exp-run given the unknowns here. Can you do as A/B test on 10.3 without/with the patch, both builds from scratch and against the same revision?
(In reply to Jan Beich (mail not working) from comment #8) Exp-run results for 11.0 amd64 are available at http://package18.nyi.freebsd.org/jail.html?mastername=110amd64-default-PR215401 Exp-run results for 11.0 i386 are available at http://package18.nyi.freebsd.org/jail.html?mastername=110amd64-default-PR215401
(In reply to Antoine Brodin from comment #9) http://package18.nyi.freebsd.org/jail.html?mastername=110i386-default-PR215401
Exp-run looks fine, review still welcome.
This should come out too: Mk/Scripts/functions.sh:180: ARCH \
Exp-run has to be done again with the Mk/Scripts/functions.sh change
A few failures that seem due to this change: http://pb2.nyi.freebsd.org/data/110amd64-default-foo/2017-01-15_09h46m04s/logs/errors/iortcw-1.5.a,1.log http://pb2.nyi.freebsd.org/data/110amd64-default-foo/2017-01-15_09h46m04s/logs/errors/legesmotus-0.4.0_5.log http://pb2.nyi.freebsd.org/data/110amd64-default-foo/2017-01-15_09h46m04s/logs/errors/warsow-1.51_2.log http://pb2.nyi.freebsd.org/data/110amd64-default-foo/2017-01-15_09h46m04s/logs/errors/zhlt-3.4.log http://pb2.nyi.freebsd.org/data/110amd64-default-foo/2017-01-15_09h46m04s/logs/errors/ja-jed-0.99.8.j067_5.log http://pb2.nyi.freebsd.org/data/110amd64-default-foo/2017-01-15_09h46m04s/logs/errors/afni-2015.11.13_1.log http://pb2.nyi.freebsd.org/data/110amd64-default-foo/2017-01-15_09h46m04s/logs/errors/consolehm-1.31_1.log
Some extra failures on i386: http://pb2.nyi.freebsd.org/data/110i386-default-foo/2017-01-19_15h44m38s/logs/errors/q3cellshading-1.0_2.log http://pb2.nyi.freebsd.org/data/110i386-default-foo/2017-01-19_15h44m38s/logs/errors/quake3-1.32c_4.log http://pb2.nyi.freebsd.org/data/110i386-default-foo/2017-01-19_15h44m38s/logs/errors/tremulous-1.1.0_11.log
batch change of PRs untouched in 2018 marked "in progress" back to open.
Moin moin Closing this, due to old age. Please re-open the issue, if it is still relevant. mfg Tobias