Created attachment 217857 [details]
A recent build-report showed a problem building the vile port.
The symptom pointed to a problem with parallel make.
The likely problem seems to be as reported against the make program here:
"make: parallel builds and double-colon don't get along"
That is, double-colon targets in the makefiles cause the build to fail.
vile uses this in several places, and rewriting the makefiles to work
around a bug in a given make-program isn't going to happen. It works
well enough with gmake, but is easy to reproduce with bmake with my
Debian/testing for instance:
--- all ---
--- bnf-filt.o ---
rm -f bnf-filt.c
--- css-filt.o ---
echo "#include <flt_defs.h>" > css-filt.c
flex -t css-filt.l >> css-filt.c
clang -c -I. -I.. -I../filters -I.. -DHAVE_CONFIG_H -D_DEFAULT_SOURCE -D_XOPEN_SOURCE=500 -D_FILE_OFFSET_BITS=64 -I./filters -g -O2 -Qunused-arguments -Wno-error=implicit-function-declaration css-filt.c
bmake: all#1: cohort has parents
bmake: stopped in /usr/build/vile/vile
The problem with make appears to be recent; it works with FreeBSD 12.
(I tried installing FreeBSD-13-current...). The Debian/testing package
also is new:
The port should be updated to suppress the parallel-make feature.
The attachment sets MAKE_JOBS_UNSAFE=yes as documented in
however... reviewing the logs, I see that even with this change
the make program (in FreeBSD 12) gets a "-j4" option when redurring
to the filters subdirectory.
The leftover "-j4" was actually from a test-script which I used for checking ports. (The suggested update to the makefile works as intended).
The patch in bug249224 is preferred, right?
no - these are two different ports for the same source.
I'm puzzled because I created the one for editors/vile first,
and the title looked correct. I fixed it now...
(In reply to Thomas E. Dickey from comment #4)
Just a few minutes ago, I inspected both attachments, and they were for the correct ports, but on revisiting, both are for editors/vile. I'll attach a diff instead.
(In reply to Thomas E. Dickey from comment #5)
When I opened this report, I copied the description from the other (editors/vile), without noticing that the copy/paste copied the link to the attachment.
Adding to the confusion (which oddly enough did not prevent building the port) was an earlier change to the portname:
Created attachment 217866 [details]
replacement for editors/xvile/Makefile