Resending another message from Ross with interesting info, my reply follows. -k
Date: Thu, 27 Feb 2025 18:30:43 +0000 From: Ross Burton <ross.bur...@arm.com> To: Karl Berry <k...@freefriends.org> Subject: Re: automake-1.17.90 pretest released [alpha] I need to eat but one quick datapoint from doing a three-way comparison with meld. In a clean build with 1.17, autom4te.cache/traces.0 says: m4_if([$1], [1.17], [], In a clean build with 1.17.90, it says m4_if([$1], [1.17.90], [], But in a build with 1.17.90 that has previously been configured with 1.17: m4_if([$1], [1.17], [], That feels like a smoking gun to me. Relatedly, the autom4te.cache/output.0 for the build with 1.17 has this line in: _am_tools='gnutar pax cpio none' With 1.17.90 from a clean source tree: _am_tools='gnutar plaintar pax cpio noneâ # The POSIX 1988 'ustar' format is defined with fixed-size fields. [ ⦠big block of code ] But 1.17.90 from a dirty tree that previously was configured with 1.17: _am_tools='gnutar pax cpio noneâ [ no block of code ] So some state from the old build is not being purged when it should be? Cheers, Ross
