At 2022-05-28T13:15:31-0500, G. Branden Robinson wrote: > I'll have to try Ralph's suggestion regarding git-version-gen's > --fallback option; I had already made one attempt, but maybe I > misunderstood or misused it.
No, I found a bug in the git-version-gen script.
It needs the following patch.
--- build-aux/git-version-gen~ 2022-06-03 18:34:08.175890973 +0000
+++ build-aux/git-version-gen 2022-06-03 17:51:02.221945195 +0000
@@ -189,7 +189,7 @@
# Remove the "g" to save a byte.
v=`echo "$v" | sed 's/-\([^-]*\)-g\([^-]*\)$/.\1-\2/'`;
v_from_git=1
-elif test "x$fallback" = x || git --version >/dev/null 2>&1; then
+elif test "x$fallback" = x && ! git --version >/dev/null 2>&1; then
v=UNKNOWN
else
v=$fallback
I've documented this in our "INSTALL.REPO" file.
I don't see how the old logic could ever have worked, so I have to
conclude that the gnulib developers did not test this scenario.
Disappointing. (Ingo's entitled to say "I told you so".)
Nevertheless, groff snapshot builds seem to work now, for the first time
ever.
I've also tested several more different build environment scenarios
(with various optional dependencies absent). Most of my recent commits
have been to make these either work, or to make them more tidy and less
productive of noisy diagnostics.
Long story short: if the groff tar archive you grab unpacks a
"configure" script, you're using a distribution archive and need the
instructions in "INSTALL.extra". If it doesn't, it's a snapshot archive
and you need the instructions in "INSTALL.REPO".
Maybe we can get more people to try snapshot builds now.
Regards,
Branden
signature.asc
Description: PGP signature
