Re: pkgsrc-2023Q2 openjdk17 failure on netbsd-9 amd64

[Paul Ripke <stix%stix.id.au@localhost>]

On Tue, Jul 04, 2023 at 04:27:31PM +1000, Paul Ripke wrote:
> On Mon, Jul 03, 2023 at 12:39:58PM +1000, Paul Ripke wrote:
> > Updating via pkg_rolling-replace (also same failure with "make stage-install"):
> > 
> > ...
> > 2023-07-03-08:37:35: Generating keytool wrapper...
> > 2023-07-03-08:37:35: Generating rmiregistry wrapper...
> > 2023-07-03-08:37:35: Generating serialver wrapper...
> > 2023-07-03-08:37:35: ===> install-message [openjdk17-1.17.0.6.10nb2] ===> Installing for openjdk17-1.17.0.6.10nb2
> > 2023-07-03-08:37:35: => Generating pre-install file lists
> > 2023-07-03-08:37:35: => Creating installation directories
> > 2023-07-03-08:37:35: /usr/bin/install -d -o stix -g users -m 755 /home/tmp/pkgwrk/lang/openjdk17/work/.destdir/usr/pkg/java/openjdk17
> > 2023-07-03-08:37:35: cd /home/tmp/pkgwrk/lang/openjdk17/work/jdk17u-jdk-17.0.6-10-1/build/bsd-x86_64-server-release/images/jdk && pax -rwpp .            /home/tmp/pkgwrk/lang/openjdk17/work/.destdir/usr/pkg/java/openjdk17/.
> > 2023-07-03-08:37:35: cd: can't cd to /home/tmp/pkgwrk/lang/openjdk17/work/jdk17u-jdk-17.0.6-10-1/build/bsd-x86_64-server-release/images/jdk
> > 2023-07-03-08:37:35: *** Error code 2
> > 2023-07-03-08:37:35:
> > 2023-07-03-08:37:35: Stop.
> > 2023-07-03-08:37:35: make[1]: stopped in /usr/pkgsrc/lang/openjdk17
> > 2023-07-03-08:37:35: *** Error code 1
> > 2023-07-03-08:37:35:
> > 2023-07-03-08:37:35: Stop.
> > 2023-07-03-08:37:35: make: stopped in /usr/pkgsrc/lang/openjdk17
> > 2023-07-03-08:37:35: *** 'make replace' failed for package openjdk17.
> > 2023-07-03-08:37:35: *** Please read the errors listed above, fix the problem,
> > 2023-07-03-08:37:35: *** then re-run pkg_rolling-replace to continue.
> > 
> > Sure enough, that directory doesn't exist; there's no "images" inside
> > "bsd-x86_64-server-release".
> > 
> > I don't appear to have any relevant changes in my tree, not anything
> > interesting in /etc/mk.conf. This pkg built fine with pkgsrc-2023Q1.
> 
> Ahh, I see the bulk builds are also failing, so it's not just me. I
> may take a look if I get a chance.

Finally figured this out. Turns out, the invocation of gmake wasn't
actually performing the full build, apparently due to parsing the
environment variable 'MAKEFLAGS'. Fix was adding:

MAKE_ENV+=      MAKEFLAGS=

For reference, it seems that UNPRIVILEGED_GROUPS somehow triggers
gmake to give up (backslashes & spaces?) - although just emptying the
variable as above seems to work.

ksh$ pwd
/usr/pkgsrc/lang/openjdk17
ksh$ make show-var VARNAME=MAKEFLAGS
_MAKE=/usr/bin/make OPSYS=NetBSD OS_VERSION=10.0_BETA OPSYS_VERSION=100000 LOWER_OPSYS=netbsd _PKGSRCDIR=/usr/pkgsrc PKGTOOLS_VERSION=20210410 _CC=/usr/bin/gcc UNPRIVILEGED_USER=stix UNPRIVILEGED_GROUP=users UNPRIVILEGED_GROUPS=users\ wheel\ operator\ nvmm\ squid\ cvsadmin _PATH_ORIG=/home/stix/bin:/home/stix/bin/NetBSD.x86_64:/sbin:/usr/sbin:/bin:/usr/bin:/usr/games:/usr/pkg/sbin:/usr/pkg/bin:/usr/local/bin:/usr/X11R7/bin:/etc:/usr/pkg/java/openjdk17/bin:/usr/pkg/py39-google-cloud-sdk/bin

Since gmake parses MAKEFLAGS, shouldn't we be doing something about
this variable in general?

-- 
Paul Ripke
"Great minds discuss ideas, average minds discuss events, small minds
 discuss people."
-- Disputed: Often attributed to Eleanor Roosevelt. 1948.