Hi!
Each iteration took its time (in the range of two to four days), and I
learned quite a lot in between. ;-)
So first of all, my goal is to create a (standalone) pbulk system as
a first step, possibly using distcc to relay some compilation work to
a simple PC. Main documentation is:
[1] http://wiki.netbsd.org/tutorials/pkgsrc/pbulk/
[2] https://wiki.netbsd.org/tutorials/pkgsrc/cross_compile_distcc/
and [3] https://hackaday.io/project/218-speed-up-pkgsrc-on-retrocomputers/details
Using [1], it always broke during
`sh /usr/pkgsrc/mk/pbulk/pbulk.sh -n -c mk.conf.frag` due to missing
`-O0` with current gcc.old that's in use. Adding this as CFLAGS into
the mk.conf.frag didn't help, and adding it to (within the sandbox)
/etc/mk.conf also resulted in no difference. Looking through the
scripts, this finally helped:
bash-5.2# pwd
/usr/pkgsrc/mk/pbulk
bash-5.2# ls -l
total 14
drwxrwxr-x 2 600 125 512 Feb 13 01:00 CVS
-rw-r--r-- 1 600 125 4708 Feb 13 01:00 pbulk-index.mk
-rw-r--r-- 1 600 125 5202 Jul 23 20:44 pbulk.sh
bash-5.2# cvs diff -u .
cvs diff: Diffing .
Index: pbulk.sh
===================================================================
RCS file: /cvsroot/pkgsrc/mk/pbulk/pbulk.sh,v
retrieving revision 1.9
diff -u -r1.9 pbulk.sh
--- pbulk.sh 2 Dec 2020 10:22:39 -0000 1.9
+++ pbulk.sh 27 Jul 2023 09:34:04 -0000
@@ -64,7 +64,8 @@
# Deploying pbulk packages:
# - bootstrapping
cat >${TMPDIR}/pbulk.mk <<EOF
-PKG_DEVELOPER= yes
+PKG_DEVELOPER= no
+CFLAGS+=-O0
EOF
${PKGSRCDIR}/bootstrap/bootstrap \
So there's a hardcoded mk.conf that the build process starts with. It
breaks first while building the digest tools (ICE), later on while
building bison (pbulk -> rsync -> bash -> bison) due to
"PKG_DEVELOPER" checks, which reveal that `awk` may have FP issues.
But finally, a `pbulk` capable environment seems to have emerged and
I'm now trying to build an initial package (sysutils/pstree). I though
guess that my <sandbox>/usr/pbulk/etc/pbulk.conf isn't all that cool,
but it's a hugh step forward.
During earlier fiddlink (which led to Python 3.10 being built), I
also learned that there are even more GCC and Python patches flying
around. These should, at least, fix the VAX-specific GCC bugs we're
all getting bitten by! I hope we can get these merged and checked wrt.
upstream GCC. That would be highly welcomed! :)
MfG, JBG
--
Attachment:
signature.asc
Description: PGP signature