Failed to emerge app-portage/portage-utils-0.87

[jhon987]

Errors start here:

[Ionen]

I am unable to reproduce on neither amd64 (fresh stage3) nor my ~amd64 systems (no gcc10 though), not seeing any bug open about this either.

Given the very recent stabilization process I'd think there's no problems on normal configurations, so `emerge --info` output may help.

[jhon987]

yeah I figured this probably uniqe to me, since I haven't seen any othr reference...

[Ionen]

Yeah, I get same error if I use your -fno-builtin-strlen which immediately looked suspicious.

Edit: I assume you may have used this for valgrind maybe? I don't think having this set system-wide is a good idea, if you have a reason for it set it only per-package as needed

[fedeliallalinea]

As Ionen say try to rebuild with standard CLFLAGS -march=native -O2 -pipe
_________________
Questions are guaranteed in life; Answers aren't.

[jhon987]

Indeed I use it for valgrind.
I confirm removing it solves the problem, thank you very much!

@Ionen

[fedeliallalinea]

[Ionen]

Well, it's up to you. You can also setup the other way around if you really prefer (as-in use package.env to re-enable the builtin per failing package).

You also "could" report it as an issue to upstream of affected packages if you want this fixed, technically packages are not "supposed" to fail with this. Although not every developers care about things like this.

Next time you have issues, do at least give it a try with normal cflags anyhow. Issue may come up again if left system-wide.

[jhon987]

@Ionen thank U for understanding.
the problem is I tend to forget what is normal cflags sometimes (see my explanation below)...

@fedeliallalinea look, here is the thing. I only program as hobby (not professional), it can be sometimes months, or more, that I haven't touch any code. And then, upon fixing some bug of a program I wrote long time ago, or for programing something small every once in a while - I simply don't remember what use flags I enabled or even how to use valgrind properly, and then I have to reread valgrind manual and so forth.
So it's much easier for me to set and forget, instead of resetting, and relearning every rare time I use that feature.

The downside, of course, is because I don't remember, then I post a question here, and that's troubling you guys to help me find solution for my problem, but then again, it's not like I'm making you do it and as Ionen said,

[Ionen]

[Hu]

In this case, the original code looks questionable to me, and could probably be made to work by using sizeof(#V) - 1 instead of strlen(#V). sizeof would be a true compile-time expression, and should be safe since it is being passed a stringized constant.

[GDH-gentoo]

Right. The code contains a "constraint violation". But upstream is also Gentoo in this case, so bug filed:

Bug #723422

[grobian]

This is correct, and should've used sizeof() - 1 regardless of this compiler behaviour.

I've pushed a fix, can someone confirm that portage-utils-9999 compiles correctly now?

Thanks
_________________
Gentoo on a different level

[Ionen]

[grobian]

right, I missed that one, made another push just now
_________________
Gentoo on a different level

[Ionen]