[M3commit] CVS Update: cm3

Jay K jay.krell at cornell.edu
Mon Aug 3 04:59:53 CEST 2009


more tangential attachments regarding the pain of getting a current gcc/g++ on some systems..
This isn't all necessary coherent/working, I stopped working on it.
I was also attempting to minimize rebuilding e.g. build libiberty when doing multiple cross builds.
 
 - Jay



----------------------------------------
> From: jay.krell at cornell.edu
> To: hosking at cs.purdue.edu; jkrell at elego.de
> Date: Mon, 3 Aug 2009 02:40:06 +0000
> CC: m3commit at elegosoft.com
> Subject: Re: [M3commit] CVS Update: cm3
>
>
> I didn't realize that. Maybe we should enable-bootstrap in some cases? (and let fixincludes/libgcc build?)
> I happen to have deliberately emptied my .profile file, to try to isolate things from any of my customizations, which leads to cc being in the default $PATH but not gcc.
> Granted, there might have a default .profile that was a litte "better".
> I don't even have ar and as in $PATH now (no /usr/ccs/bin, jus /usr/bin).
>
> Or maybe user is responsible for having a new enough gcc, and in $PATH, and m3makefile should just
> if not equal(try_exec("type gcc 2>/dev/null", 0))
> error("no gcc in $PATH")
> end
> if not equal(try_exec("type ar 2>/dev/null", 0))
> error("no ar in $PATH")
> end
>
> and be done? Granted, it could also cc -v | grep -i gnu or something.
>
>
> You know..sometimes gcc isn't so easy to build from a bare system with vendor cc.
> (see attached rigamarole I was working on for Irix and AIX; HP-UX is also painful, since the bundled compiler is only K&R and you have o first use it to build gcc 3.3 or 3.4, goal being to get up to a current gcc, not just an adequate gcc for cm3cg; you run into problems like vendor make can't build out of tree, so you have to build GNU make; vendor tar can't extract the full gcc tree due to @LongLink, so you have to build GNU tar, but vendor cc can' build GNU tar, so you build just gcc-core first, and also that vendor cc slightly miscompiles GNU make (never tracked that down) such that the first gcc is only partly working, use that to rebuild make and then rebuild gcc-core, and then tar, and then full gcc, something like that...)
>
>
>
>
> - Jay
>
>
> ________________________________
>> From: hosking at cs.purdue.edu
>> To: jkrell at elego.de
>> Date: Sun, 2 Aug 2009 20:29:10 -0400
>> CC: m3commit at elegosoft.com
>> Subject: Re: [M3commit] CVS Update: cm3
>>
>> I'm not sure I like these "fixes" to gcc just for compiling with non-gcc C compilers. gcc is always intended to be bootstrapped by itself, though often the phase 1 bootstrap comes from compiling with the native cc.
>>
>> On 3 Aug 2009, at 00:40, Jay Krell wrote:
>>
>> CVSROOT: /usr/cvs
>> Changes by: jkrell at birch. 09/08/03 00:40:15
>>
>> Modified files:
>> cm3/m3-sys/m3cc/gcc/gcc/: tree-ssa-loop-ivopts.c
>>
>> Log message:
>> fix for compiling with Sun cc (also fixed in gcc trunk)
>>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: build2.py
URL: <http://m3lists.elegosoft.com/pipermail/m3commit/attachments/20090803/33fac167/attachment-0004.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: build.py
URL: <http://m3lists.elegosoft.com/pipermail/m3commit/attachments/20090803/33fac167/attachment-0005.ksh>


More information about the M3commit mailing list