lha: fix Xcode 12 build issues

[mitchblank]

There are actually two separate problems that got triggered by Xcode 12's decision to make -Werror,-Wimplicit-function-declaration the default:

• The ./configure script has problems. Some of these can be fixed with an autoreconf run, but the non-HEAD version of the code needs other adjustments to work with modern autoconf. However there are a couple other problems which I opened an upstream PR for already: Fix autoconf issues with Xcode 12 jca02266/lha#18
• There is also a build issue. macports already has a fix for that: lha: fix implicit function declarations macports/macports-ports#9197
  However the simplest thing to do for the moment is to just disable the warning via $CFLAGS which allows the code to build as-is

[ryandesign]

• the simplest thing to do for the moment is to just disable the warning via $CFLAGS which allows the code to build as-is

but possibly be wrong or crash on Apple Silicon. (Why do you think Apple made this an error in the first place?) Better to fix it properly.

[mitchblank]

but possibly be wrong or crash on Apple Silicon. (Why do you think Apple made this an error in the first place?) Better to fix it properly.

We're dealing with a lot of Xcode 12 failures that are due to this error change (you can look through my recent PRs for a bunch of them, and I'm not the only person who has been working on this) The most typical places this happens is when some simple code -- often in an autoconf fragment -- does something as simple as call exit(0); without having done #ifdef <stdlib.h> There is little risk that this would do the wrong thing.

I 100% agree with you that the best fix is to make a real fix which is why I have been upstream ideal fixes whenever I can. In this case I assume that you are on top of it for your macports change; as you can see I've opened a PR on upstream for the configure issue I diagnosed. However, until the next upstream release I think the $CFLAGS workaround is fine.

WRT your macports change, how risky do you see that one -- I don't know enough about the apple ARM ABI to know if the implicit function signature is likely to cause a mismatch there.

[ryandesign]

My understanding is that the ARM machine code for calling a function is different depending on whether the function is variadic or non-variadic. The compiler can't know whether the function is variadic or non-variadic without seeing the function declaration. If it guesses wrong, the program will crash. I assume it would guess that the function is non-variadic, and since exit is non-variadic, suppressing the error would probably be ok in this case.

[BrewTestBot]

🤖 A scheduled task has triggered a merge.