[Bug 593] gcc/config/t-rtems + gcc/config/<cpu>/t-rtems inclusion order

Tue Aug 5 17:00:33 UTC 2008

http://www.rtems.org/bugzilla/show_bug.cgi?id=593

Joel Sherrill <joel.sherrill at oarcorp.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|tools                       |GCC
            Product|RTEMS                       |Tools
           Platform|                            |All
            Version|unknown                     |unspecified

The order of including the t-rtems files in gcc's config.gcc is inconsistent.

Some include config/t-rtems before config/<cpu>/t-rtems, other apply the opposite order.

Release:
N/A

--- Comment #1 from Ralf Corsepius <ralf.corsepius at rtems.org>  2004-03-17 11:31:01 ---
Fix:
We should reach an agreement about the correct inclusion order  and change config.gcc accordingly.

Core question: Is t-rtems supposed to provide defaults to be overridden by <cpu>/t-rtems or is t-rtems supposed to override setting from <cpu>/t-rtems?

I don't know the answer off-hand. At least until gcc-3.3 both cases were required for particular targets.

May-be we should we split t-rtems into a t-rtemspre and t-rtemspost? I am not sure, but I don't think this will be necessary.

--- Comment #2 from Ralf Corsepius <ralf.corsepius at rtems.org>  2004-03-18 14:04:01 ---
From: Ralf Corsepius <ralf_corsepius at rtems.org>
To: Chris Johns <cjohns at cybertec.com.au>
Cc: RTEMS Bugs <bugs at rtems.com>
Subject: Re: tools/593: GNATS-RTEMS-gcc/config/t-rtems +
	gcc/config/<cpu>/t-rtems inclusion order
Date: Thu, 18 Mar 2004 14:04:01 +0100

 On Thu, 2004-03-18 at 13:00, Chris Johns wrote:
 > ralf_corsepius at rtems.org wrote:
 > > 
 > > We should reach an agreement about the correct inclusion order  and change config.gcc accordingly.
 > > 
 > > Core question: Is t-rtems supposed to provide defaults to be overridden by <cpu>/t-rtems or is t-rtems supposed to override setting from <cpu>/t-rtems?
 > > 
 > > I don't know the answer off-hand. At least until gcc-3.3 both cases were required for particular targets.
 > > 
 > > May-be we should we split t-rtems into a t-rtemspre and t-rtemspost? I am not sure, but I don't think this will be necessary.
 > > 
 > 
 > What is being controlled by these file ?
 The t-* files are makefile-fragments, being used in gcc/config.gcc to
 compose parts of the "target-makefiles" (tmake_file=... in
 gcc/config.gcc).

 ATM,
 * gcc/config/t-rtems presets/hard-codes some very fundamental general
 make-variables, being used by building all *-rtems* targets. 

 * The gcc/config/<cpu>/t-rtems files do the same, but on a per-cpu
 basis. ATM, they essentially are used to override the default set of
 multilibs with RTEMS-specific multilibs.

 In an ideal world all CPUs being supported by RTEMS would have a
 (possibly empty) <cpu>/t-rtems and would be using the same order to of
 t-rtems <cpu>/t-rtems in their corresponding tmake_file-line in
 config.gcc.

 ATM, the situation is as follows (gcc-3.4-branch):
 # grep t-rtems config.gcc
         tmake_file="arm/t-arm-elf t-rtems"
         tmake_file="c4x/t-c4x t-rtems"
         tmake_file="h8300/t-h8300 t-rtems h8300/t-rtems"
         tmake_file="pa/t-pro t-rtems"
         tmake_file="i386/t-rtems-i386 i386/t-crtstuff t-rtems"
         tmake_file="i960/t-960bare t-rtems"
         tmake_file="m68k/t-m68kbare m68k/t-crtstuff t-rtems m68k/t-rtems"
         tmake_file="mips/t-elf t-rtems mips/t-rtems"
         tmake_file="rs6000/t-fprules rs6000/t-rtems t-rtems rs6000/t-ppccomm"
         tmake_file="sh/t-sh t-rtems"
         tmake_file="sh/t-sh sh/t-elf t-rtems"
         tmake_file="sparc/t-elf sparc/t-crtfm t-rtems"

 => 
 Some cpus use "t-rtems <cpu>/t-rtems", 
 some cpus use "<cpu>/t-rtems t-rtems",
 some cpus have a <cpu>/t-rtems, many don't,
 some cpus use "<cpu>/t-rtems<something" some use "<cpu>/t-rtems"

 In one sentence: "It's a scheme having historically evolved, without
 much systematic in it".

 Ralf

--- Comment #3 from Chris Johns <chrisj at rtems.org>  2004-03-18 23:00:30 ---
From: Chris Johns <cjohns at cybertec.com.au>
To: ralf_corsepius at rtems.org
Cc: bugs at rtems.com
Subject: Re: tools/593: GNATS-RTEMS-gcc/config/t-rtems + gcc/config/<cpu>/t-rtems
 inclusion order
Date: Thu, 18 Mar 2004 23:00:30 +1100

 ralf_corsepius at rtems.org wrote:
 > 
 > We should reach an agreement about the correct inclusion order  and change config.gcc accordingly.
 > 
 > Core question: Is t-rtems supposed to provide defaults to be overridden by <cpu>/t-rtems or is t-rtems supposed to override setting from <cpu>/t-rtems?
 > 
 > I don't know the answer off-hand. At least until gcc-3.3 both cases were required for particular targets.
 > 
 > May-be we should we split t-rtems into a t-rtemspre and t-rtemspost? I am not sure, but I don't think this will be necessary.
 > 

 What is being controlled by these file ?

 -- 
   Chris Johns

--- Comment #4 from Joel Sherrill <joel.sherrill at oarcorp.com>  2008-08-05 12:00:33 ---
Moving to Tools

-- 
Configure bugmail: http://www.rtems.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.