Fix issue 19116 - Segfaults on recent Linux 32 bits distros

[SSoulaimane]

Pass the -no-pie flag to cc on Linux 32 bits.

Re-enables CircleCi 32 bits tests which were previously disabled in #9499.

[WalterBright]

I think it would be much better to find out what the PIE code gen issue is.

[WalterBright]

https://issues.dlang.org/show_bug.cgi?id=19116

[SSoulaimane]

@WalterBright There is no codegen issue. LDC turned on PIC by default ldc-developers/ldc#1618 & ldc-developers/ldc#1664

[SSoulaimane]

The rust compiler runs the linker with -no-pie if it fails it removes it. rust-lang/rust@f0e9af1.

[wilzbach]

@SSoulaimane uhm thanks a lot for this work, but I'm not sure this is a good investment of your time. As this has been broken for years, it clearly shows that either no one uses DMD 32-bit anymore on Linux or they switched to LDC. In either case, the 32-bit -no-pie test is already done by the auto-tester, so there's no real gain of this (except for adding more load to the circleci build queue).

[SSoulaimane]

@wilzbach Thanks. It's not a hard problem though. We just need to make a decision, we either turn PIC on, or read the linker output.

[PetarKirov]

@SSoulaimane I suggest to turn PIC on by default and provide the option to disable it for the rare cases where this may be necessary.

[SSoulaimane]

@PetarKirov I thought about it but then if we had a working option that turns PIC off we could just use it and keep PIC off by default.
I found out that there is an easy reliable way to probe the compiler and see whether it has this default PIE or not. Just run cc -v and see if --enable-default-pie is in the output.

[SSoulaimane]

@wilzbach I thought only 32bits had problems but turning on CircleCi no_pic tests revealed that x86_64 tests fail when linking phobos (see link) but the -no-pie flag fixes it. Is no_pic no longer supported on x86_64?

[wilzbach]

Is no_pic no longer supported on x86_64?

Uhm I'm not sure how I am supposed to know this, though most/all? major Linux distributions have switched to PIC only quite a while ago, so I wouldn't be surprised if non-PIC is no longer tested nor supported unofficially. There would be little point in fixing it in this case.

[SSoulaimane]

@wilzbach It's fixed now I don't know if you want to keep the circle ci tests or not.

[dlang-bot]

Thanks for your pull request and interest in making D better, @SSoulaimane! We are looking forward to reviewing it, and you should be hearing from a maintainer soon.
Please verify that your PR follows this checklist:

• My PR is fully covered with tests (you can see the coverage diff by visiting the details link of the codecov check)
• My PR is as minimal as possible (smaller, focused PRs are easier to review than big ones)
• I have provided a detailed rationale explaining my changes
• New or modified functions have Ddoc comments (with Params: and Returns:)

Please see CONTRIBUTING.md for more information.

---

If you have addressed all reviews or aren't sure how to proceed, don't hesitate to ping us with a simple comment.

Bugzilla references

Auto-close	Bugzilla	Severity	Description
✓	19116	normal	dmd compiles binary but segfaults immediately when run on Ubuntu / Debian 32Bit architectures

Testing this PR locally

If you don't have a local development environment setup, you can use Digger to test this PR:

dub fetch digger
dub run digger -- build "master + dmd#9981"

[wilzbach]

What's the overhead in opening this process on every compilation? 🤔

[SSoulaimane]

The linker is slow anyway would be an argument.
But maybe we can pessimize older cc versions instead of the new ones i.e optimistically assume that cc can handle the -no-pie switch and retry if it fails.