tinycc-devel
[Top][All Lists]
Advanced

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

Re: [Tinycc-devel] can't build mob branch on x86 64


From: Thomas Preud'homme
Subject: Re: [Tinycc-devel] can't build mob branch on x86 64
Date: Fri, 4 Feb 2011 20:52:23 +0100

Le vendredi 04 février 2011 20:45:51, Thomas Preud'homme a écrit :
> Le vendredi 04 février 2011 20:23:13, grischka a écrit :
> > tcctok.h:
> >  > #if defined __i386__ || defined __x86_64__
> >  > 
> >  >      DEF(TOK_alloca, "alloca")
> >  > 
> >  > #endif
> > 
> > Not all compilers do define __i386__ or __x86_64__, e.g. MSVC
> > does not.  (It defines _X86_ and _AMD64_).
> > 
> > Also, the host platform doesn't matter anyway, the question
> > here is whether alloca is defined for the target platform.
> 
> Yes but how to detect that alloca is available to define TOK_alloca?
> 
> > In test/tcctest.c:
> >  > void alloca_test()
> >  > {
> >  > #ifdef TOK_alloca
> > 
> > tcctest.c is not part of the tcc source code and hence never
> > defines "TOK_alloca" (and ahould not).
> 
> So same problem, previously only __i386__ and __x86_64__ were tested. Do
> you have an advice for that? I guess for the test we could detect at
> runtime thanks to dlopen/dlsym but it seems like overkill solution.

I meant the result of dlopen(RTLD_DEFAULT, "alloca"); could be used, that 
would not be so big code after all, and as the test are very quick to run it's 
not really a problem for performances. But I still don't know how to define 
TOK_alloca conditionally.
> 
> > Anyway, nice work.  I guess this could be a good headline for next
> > release: "tcc 0.9.26 now supports C99 type VLAs." ;)
> 
> Thanks, take it as nice compliment from you but when I see the size of the
> patch I know it's really nothing. Anyway it has several times been asked,
> and I wasn't happy with the solution we used in Debian. It was taken from
> the forge BTS and it was basically rewriting the source code and handled
> only one level of vla.
> 
> By the way, there is a couple of bugs that could be closed in the savannah
> BTS, namely #30457 [1] , #23851 [2], #13213 [3]. Some other bugs might also
> be closed as well I think, like #15366 [4].
> 
> > --- grischka
> 
> Best regards,
> 
> Thomas Preud'homme
> 
> [1] http://savannah.nongnu.org/bugs/?30457
> [2] http://savannah.nongnu.org/bugs/?23851
> [3] http://savannah.nongnu.org/bugs/?13213
> [4] http://savannah.nongnu.org/bugs/?15366

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


reply via email to

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