tinycc-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Tinycc-devel] test failure on Linux


From: Thomas Preud'homme
Subject: Re: [Tinycc-devel] test failure on Linux
Date: Thu, 14 Feb 2013 15:45:46 +0100
User-agent: KMail/1.13.7 (Linux/3.2.0-4-amd64; KDE/4.8.4; x86_64; ; )

Le jeudi 14 février 2013 12:52:13, grischka a écrit :
> Thomas Preud'homme wrote:
> > Oh crap. Can you tell me whether there is still a libc.so.6 in /lib?
> > 
> > If yes, then I could replace the ldd call by something like ls
> > /lib/**/libc.so.6 (of course it would not be ** since it's not bourne
> > shell)
> 
> Don't know right now but it works fine without anything set,
> just from the defaults in tcc.h.  Anyway I'd suggest to stay
> conservative for now.  That is to make a more or less static
> assumption about mulitarch.  For example you could add the code
> that you had in tcc.h previously, into conftest.c, such as
> 
>      case 't':
>          printf("TRIPLET_ARCH "-" TRIPLET_OS "-" TRIPLET_ABI\n");
>       break;
> 
> Or you do still something with ldd and sed but it's risky.

This solution is not very pretty but as you suggested, it's better to be safe. 
I was thinking about looking at what ld.so is used, then readlink -e to get 
the full path and then do the same logic. That'll wait for after the release 
then.

> 
> In any case, then check whether the guess does make sense, for
> example whether there is
>       /usr/lib/guessed-MUA/crt1.o.
> 
> If not, just do nothing.  Exept on old x86_64, you could
> set lddir=lib64 if it exists.

Yep, just after sending my mail about finding libc.so.6 I thought it might be 
better to look for crti.o. I hope that approach will please you.

> 
> Previously we handled this cases in Makefile:
> 
>       "i386-linux-gnu"
>       "i386-kfreebsd-gnu"
>       "i386-gnu"
>       "lib64"
>       "x86_64-linux-gnu"
>       "x86_64-kfreebsd-gnu"
>       "arm-linux-gnueabi"
>       "arm-linux-gnueabihf"
> 
> Which seemed to have worked for everyone except for the empty
> MUA-dir case from the OP (Ramsay).

Hopefully, now it'll work for everyone. Basically it's the same thing, we 
don't detect the triplet but instead deduce it from the build machine and then 
check wether there is a crti.o at this place.

> 
> --- grischka

Best regards,

Thomas

Attachment: signature.asc
Description: This is a digitally signed message part.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]