Round 2 of my and Petr Vorel's patch set to polish up config.guess
a little. The first two patches are the most important, as they
future-proof config.guess against new C compilers that object to
implicit int and/or old-style function definitions.
The second two attempt to address a user experience problem where
people don't realize that config.guess is choking on their system
because they're trying to build a 20-year-old tarball on a CPU
architecture that didn't exist when that tarball was created.
They have incidental textual dependencies on the first two patches
but are otherwise unrelated.
I do not have access to any of the really old systems affected by
these changes, so careful review would be appreciated. In particular
patch 4/4 may not do the right thing in cases where the big switch
falls through without setting $GUESS.
Also, for additional future-proofing I'd like to create HTTP redirects
at www.gnu.org/software/config.{guess,sub} that point to Savannah CGIT,
so that the URLs baked into config.guess can be stable even if
the VCS for these files changes again (as it has in the past).
How do we make that happen?
Petr Vorel (1):
config.guess: Avoid implicit int in definitions of main.
Zack Weinberg (3):
config.guess: Modernize C programs a bit more.
config.guess: Revise diagnostic output produced on failure.
config.guess: Improve behavior when uname output is not understood.
config.guess | 251 ++++++++++++++++++++++++++++++---------------------
1 file changed, 148 insertions(+), 103 deletions(-)
--
2.43.2
