Homebrew-core: swipl-ld is broken in swi-prolog 8.0.2 (and --HEAD) recipe

[soli]

I already mentioned this issue there https://github.com/Homebrew/homebrew-core/pull/37993#issuecomment-477540922 but I guess it makes sense to open a proper issue about this.

So, to quote myself:
Though the swipl binary is correctly built by CMake now, the swipl-ld one is broken. Indeed, it stores at compile time the name of the compiler that build it (accessed from C_CC and C_CXX https://github.com/SWI-Prolog/swipl/blob/master/src/swipl-ld.c#L139), which results in it trying to call /usr/local/Homebrew/Library/Homebrew/shims/mac/super/clang at runtime, when it should only call /usr/bin/clang.

This can be checked by simply running swipl-ld -vh.
Which, on my two 10.14 macs results in:

clang: The build tool has reset ENV; --env=std required.
/usr/local/Homebrew/Library/Homebrew/shims/mac/super/clang returned code 256
*** /usr/local/Cellar/swi-prolog/HEAD-fba80fd/libexec/bin/swipl-ld exit status 1

I don't know anything about how to _not use_ shims when building swipl-ld but if that's at all possible, it would be very nice. Otherwise, overriding the C_CC and C_CXX environment variables might be enough…

[SMillerDev]

Could you fill out the issue template please? It's pretty essential in debugging issues. This issue can be reopened once we have all the information needed to start debugging.

[soli]

Oops, sorry… here it is:

• [x] are reporting a bug others will be able to reproduce and not asking a question or requesting software.
• [x] have a problem with brew install (or upgrade, reinstall) a single, Homebrew/homebrew-core formula (not cask) on macOS?
• [x] ran brew update and can still reproduce the problem?
• [x] ran brew doctor, fixed all issues and can still reproduce the problem?
• [x] ran brew gist-logs <formula> (where <formula> is the name of the formula that failed) and included the output link?
• [ ] if brew gist-logs didn't work: ran brew config and brew doctor and included their output with your issue?

And here is the brew gist-logs swi-prolog link: https://gist.github.com/soli/3b5a1c0e04bbc299f05a3d30e7b09076

[delsonsun]

Unfortunately, due to an unexpected error this message did not reach the intended recipients. Please resubmit the message.

Best regards,
EF Global IT

From: Sean Molenaar [e-mail]
Sent: Monday, April 1, 2019 9:53 AM
To: Homebrew/homebrew-core [e-mail]
Cc: Subscribed [e-mail]
Subject: Re: [Homebrew/homebrew-core] swipl-ld is broken in swi-prolog 8.0.2 (and --HEAD) recipe (#38571)

Reopened #38571https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FHomebrew%2Fhomebrew-core%2Fissues%2F38571&data=02%7C01%7C%7C5503a3b34c4e4316b44808d6b67f8339%7Cf0d1c6fddff0486a8e91cfefefc7d98d%7C0%7C0%7C636897056106583531&sdata=ZyjheATRP%2B4FP2JzWPWDavAOfTHeLCvFuftda1LIGCc%3D&reserved=0.

-
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHubhttps://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FHomebrew%2Fhomebrew-core%2Fissues%2F38571%23event-2242589169&data=02%7C01%7C%7C5503a3b34c4e4316b44808d6b67f8339%7Cf0d1c6fddff0486a8e91cfefefc7d98d%7C0%7C0%7C636897056106583531&sdata=8mzyrPXzp0K4Y%2B0wVdLw0SrjIdRBJotZ0vqixf7cIqE%3D&reserved=0, or mute the threadhttps://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAsgmqCxpRleSc9FrML3TBfZ6zMZwSXOhks5vcckIgaJpZM4cVIs7&data=02%7C01%7C%7C5503a3b34c4e4316b44808d6b67f8339%7Cf0d1c6fddff0486a8e91cfefefc7d98d%7C0%7C0%7C636897056106593541&sdata=Se3wfD2iiHvIwNsHpQ%2FkpLlCHDzIlfBonMtIL5KRPaI%3D&reserved=0.

[soli]

Since I don't see many answers, could anyone point me to some kind of docs allowing me to do the last thing I mentioned:

I don't know anything about how to not use shims when building swipl-ld but if that's at all possible, it would be very nice. Otherwise, overriding the C_CC and C_CXX environment variables might be enough…

i.e., how to change the build environment, especially related to shims.

Thanks

[MikeMcQuaid]

I don't know anything about how to _not use_ shims when building swipl-ld but if that's at all possible, it would be very nice. Otherwise, overriding the C_CC and C_CXX environment variables might be enough…

We can't avoid using the shims to build the tool but we could/should configure it so at runtime it uses the system /usr/bin/clang or similar.

[soli]

As far as I understand the source (https://github.com/SWI-Prolog/swipl/blob/master/src/swipl-ld.c) it seems that overriding $C_CC and $C_CXX at build-time might be enough, but then again I don't know how to find the "unshimmed" values nor how to change the build environment (sorry, I've never written any recipe 😭)

[sjackman]

I don't know how to find the "unshimmed" values

For git and svn:

$ $(brew --repo)/Library/Homebrew/shims/mac/super/git --homebrew=print-path
/usr/bin/git

For everything else:

$ xcrun -find clang
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang

[soli]

Ok, thanks @sjackman, I'll give a try at patching the recipe and see if I can come up with a pull request if everything works fine.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.

[soli]

If I use xcrun -find clang to set-up CMAKE_C_COMPILER (that gets put into C_CC) then when using the tool, clang does not find basic header files like stdio.h and such…

Is there an issue (I guess so, but TBH have no idea) in hardcoding /usr/bin/clang and such?

[sjackman]

Does setting C_CC=/usr/bin/clang C_CXX=/usr/bin/clang++ work as you suggested?

[soli]

Almost… swipl-ld is at least usable, but in one of the scenarios I use it, it does spit an error that I still have to investigate (I'll repost here when I'm fully ok with the result, but I wanted to know if that direction made any sense at all).

BTW, I had to set CMAKE_C_COMPILER to /usr/bin/clang (and same for CXX and clang++) as actually C_CC and C_CXX are variables that are overwritten in Params.cmake with those values…

[soli]

Ok, I managed to get rid of the error I had by adding -lc++ in some linking phases… But the resulting binary, compiled without issues, _does not work_ (it fails silently), so I'm still trying to figure out what's wrong (since in other examples now everything works well).

[sjackman]

Ok, I managed to get rid of the error I had by adding -lc++ in some linking phases…

Adding -lc++ shouldn't be necessary as long as you're linking using clang++ rather than clang or ld.

[soli]

Thanks, I know, but the point of all this issue is that I'm linking with swipl-ld (which apparently in that case does not properly detect that it should link against the C++ libraries)… So, maybe the fact that I _forced_ the C and C++ compilers at compile-time did change something in the resulting swipl-ld binary… I'm still trying to wrap my head around all this…

[sjackman]

Have you tried asking on the SWI-Prolog GitHub?
https://github.com/SWI-Prolog/swipl
https://github.com/SWI-Prolog/swipl-devel

[soli]

Not yet since I first thought it was a simple issue of using the Homebrew shims at compile time, but since it appears to be more deep, I probably will ^^

[soli]

Ok, actually this might just be related to some recent changes on SWI-Prolog's side, I've updated the Makefiles of my different projects accordingly and everything now seems to work !!! 👍
I'll have a few coworkers try my patched recipe and if it goes smoothly I'll send a PR 😎

[sjackman]

I'm glad to hear that you found a fix! That's great!

[soli]

thx for the help !