Hi Frederic - it would be better to send failures to bug-automake, so
they get bug numbers. But no big deal.
I'm facing a strange regression between 1.17.92 and 1.18 when
rebuilding Automake:
FAIL: t/txinfo-vtexi4
Are you setting SOURCE_DATE_EPOCH in the failing build and if so, to what?
Can you view the failing foo.texi file that gets created?
This output line you showed:
cat ./foo.info | grep 'GREPDATE=19 June 2025='
Makes me think the test is expecting the current date, but if
SOURCE_DATE_EPOCH is set, that won't be the case. Apparently no one
else, including me, ever ran all the tests before 1.18 with non-current
SOURCE_DATE_EPOCH :(.
Anyway, it looks to me now like the txinfo-vtexi4 test needs to be
updated in the face of mdate-sh's new behavior, i.e., it's a bug in the
test.
I'll take a look at tweaking the test asap. I surmise the shell code at
the top setting $day etc. has to be changed to use $SOURCE_DATE_EPOCH if
it's set. Which won't be fun, but so it goes. --thanks, karl.
