Hi there:
I there appears to be a race condition in the
components/Makefile line 84, between the repo creation and the
packagelint cache generation:
A .wait or .NOPARALLEL: setup would solve the race, but a dependency
between the two targets would probably be more sensible. I've tested
that it
solves this minor issue (yep, this is nit picking, but principle of
least suprise I guess :-)
What do people think to this basic patch?
diff -r 457586ba0add components/Makefile
--- a/components/Makefile Mon Aug 13 15:38:36 2012 -0700
+++ b/components/Makefile Tue Aug 14 11:48:15 2012 +0000
@@ -80,6 +80,7 @@
@cd ../tools ; echo "clobbering tools..." ; $(GMAKE) clobber
$(RM) -r $(WS_REPO) $(WS_LOGS) $(WS_LINT_CACHE)
+$(WS_LINT_CACHE): $(WS_REPO)
there's already a dependency for it on $(WS_LOGS), you could add this
one there.
setup: $(WS_LOGS) $(WS_REPO) tools $(WS_LINT_CACHE)
If its worth fixing I'll CR it and patch it in both gates (12 and
11u1) ?
If you've found a bug, sure, fix it - though I must admit I don't see
what the dependency solves.
I can see why the cache depends on the log directory, because that's
where we dump
the output, but I don't see in the Makefile where it uses the repo in
your workspace at all?
are you sure it's not something else failing? what does the log file
it creates say when it
fails (which , when it fails for me, usually is upset about reaching
ipkg :)
Am presently digging - I use -j32, and on both SPARC T4 and X86 it fails
consistently, until either the dependecy is put in or -j not specified :-)
I'll get more details for you asap.
Cheers,
Edwin
_______________________________________________
userland-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/userland-discuss
