Download raw body.
[update] editors/vis 0.8 -> 0.9
> > May I ask how are you running the tests? Just 'make test' from the ports
> > directory as usual? Additionally, does the processes started by the
> > test.sh script not inherit the shell it is executed with?
>
> Yes, "make test" from the ports dir. Might have been "make FETCH_PACKAGES=-Dsnap test".
>
> I just tried again with your new patch:
> make clean
> make
> make install (give doas my password)
> make test
> Same result, as far as I can tell. I attached vis-0.9/test.log from
> my WRKOBJDIR.
>
> My /usr/ports is managed by got, and is currently at commit 419d15f5,
> from May 3 at 16:00 UTC.
>
> > $ cat /usr/ports/pobj/vis-0.9/vis-0.9/test/sam/test.sh | head -1
> > #!/bin/sh
>
> Yes (but my WRKOBJDIR is /home/pobj):
> $ cat /home/pobj/vis-0.9/vis-0.9/test/sam/test.sh | head -1
> #!/bin/sh
A bit more info:
I get the same result on my laptop. (The bug I mentioned in my
previous email was a broken version of dwz; pkg_add -u dwz fixed
it.)
If I print the value of SHELL in sam/test.sh, or in the test rule
of vis's top-level make-file, I see an empty string.
That is: I edit /usr/ports/pobj/vis-0.9/vis-0.9/test/Makefile so it begins with:
test:
sh -c 'echo shell _$${SHELL}_'
@$(MAKE) -C core
@$(MAKE) -C lua
@$(MAKE) -C vis
@$(MAKE) -C sam
@$(MAKE) -C vim
and then when I run "make test" under /usr/ports/editors/vis the beginning of the output is:
$ make test
===> Regression tests for vis-0.9
[ -e test/Makefile ] || gmake test-update
gmake[1]: Entering directory '/usr/ports/pobj/vis-0.9/vis-0.9/test'
sh -c 'echo shell _${SHELL}_'
shell __
(As far as I can tell, putting #!/bin/sh at the start of a script
doesn't affect the value of SHELL when the script is run.)
make show=SHELL still shows /bin/sh on this machine.
--
James
[update] editors/vis 0.8 -> 0.9