* Re: Unsuccessful build of gcc-2.95.1
@ 1999-09-02 19:35 N8TM
1999-09-30 18:02 ` N8TM
0 siblings, 1 reply; 28+ messages in thread
From: N8TM @ 1999-09-02 19:35 UTC (permalink / raw)
To: gml4410; +Cc: gcc, beyerj
In a message dated 9/2/99 8:58:33 AM EST, gml4410@ggr.co.uk writes:
> I tried adding -mips3 to CFLAGS
CFLAGS is used for the gcc you are building, not for the compiler you are
using to bootstrap. More likely to work would be adding in the environment
variable e.g. CC='cc -mips3....'
Tim
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 19:35 Unsuccessful build of gcc-2.95.1 N8TM
@ 1999-09-30 18:02 ` N8TM
0 siblings, 0 replies; 28+ messages in thread
From: N8TM @ 1999-09-30 18:02 UTC (permalink / raw)
To: gml4410; +Cc: gcc, beyerj
In a message dated 9/2/99 8:58:33 AM EST, gml4410@ggr.co.uk writes:
> I tried adding -mips3 to CFLAGS
CFLAGS is used for the gcc you are building, not for the compiler you are
using to bootstrap. More likely to work would be adding in the environment
variable e.g. CC='cc -mips3....'
Tim
^ permalink raw reply [flat|nested] 28+ messages in thread
* Unsuccessful build of gcc-2.95.1
1999-09-02 9:19 Anna Winkler
@ 1999-09-30 18:02 ` Anna Winkler
0 siblings, 0 replies; 28+ messages in thread
From: Anna Winkler @ 1999-09-30 18:02 UTC (permalink / raw)
To: GCC List; +Cc: gml4410
> Date: Wed, 01 Sep 1999 19:40:43 +0100
> From: Lack Mr G M <gml4410@ggr.co.uk>
> To: gcc@gcc.gnu.org
> Subject: Unsuccessful build of gcc-2.95.1 (more of them).
>
> Is it just me???
>
> Having edited the Makefile to set "CC = cc -32" for mips-sgi-irix6.5
> the build bombs out with:
>
> > ld32: FATAL 12: Expecting n32 objects:
> > ../libiberty/libiberty.a(pexecute.o) is o32.
>
> so something is trying to use non-"-32" anyway...
You want to use -n32 instead of -32. -32 and -o32 mean generate an old
32-bit object, while -n32 is a new 32-bit object.
--Anna Winkler
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-01 17:57 N8TM
1999-09-02 4:33 ` Lack Mr G M
@ 1999-09-30 18:02 ` N8TM
1 sibling, 0 replies; 28+ messages in thread
From: N8TM @ 1999-09-30 18:02 UTC (permalink / raw)
To: gml4410, gcc
In a message dated 9/1/99 1:16:54 PM EST, gml4410@ggr.co.uk writes:
> This is for mips-sgi-irix6.5 (claimed successful).
>
> The problem is in the configure stage.
>
> The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64 ABI
> problems).
The lazy way is to set the environment variable CC='cc -n32' while using
MipsPro cc as a bootstrap compiler. AFAIK this works for MipsPro versions
since 7.2. Perhaps it's time to change a default configuration which caters
for the time before there was a working MipsPro compiler for n32/64. The
shoe probably is on the other foot now and 'cc -o32' may be less reliable.
Tim
tprince@computer.org
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 7:09 ` James Beyer
@ 1999-09-30 18:02 ` James Beyer
0 siblings, 0 replies; 28+ messages in thread
From: James Beyer @ 1999-09-30 18:02 UTC (permalink / raw)
To: Lack Mr G M; +Cc: N8TM, gcc
On Thu, 2 Sep 1999, Lack Mr G M wrote:
> > The stage1/xgcc is producing N32 mips-4 output, which fails to work
> > when using the "cc -32" option in the mh-irix6 stub. Mind you, I've
> > just noticed you suggest using "cc -n32"
>
> Ok, using cc -n32 lets it run through, but it ends up producing a
> mips4 binary, which isn't much use on an Indy, or a Challenge-XL. And
> I;m of the opinion that I shoudl be able to have one installation, and
> tell it what I want.
>
> So, I tried adding -mips3 to CFLAGS, That produces some warnign that
> indicate it hasn't worked at stage1:
>
> >> ld32: WARNING 158: Expecting MIPS3 objects: stage1/crtbegin.o is MIPS4.
Try using cc -n32 -mips3. Just using -n32 on an ORIGIN will default to
-mips4. Whereas a the other generates code that runs on my INDY.
>
> some warnings which didn't occur without the -mips3
>
> >> /usr/include/string.h:67: warning: conflicting types for
> >> built-in function `memcpy'
>
This is because of gcc's built-ins, and is something I have learned to
deal with in alot of my code. If a -fno-builtin is given to gcc it will
go away. Although I would prefer a way to quelch the message and have gcc
use its builtin calls if they truely are better.
Anyone have a better answer here?
> and then the build core dumped with a Bus error.
>
> >> ukwsv100*[1] file core
> >> core: IRIX N32 core dump of 'gengenrtl'
>
> The compilation options for this are attached.
I hope someone else knows something here!
^ permalink raw reply [flat|nested] 28+ messages in thread
* RE: Unsuccessful build of gcc-2.95.1
1999-09-02 16:10 Billinghurst, David (RTD)
1999-09-03 2:29 ` Lack Mr G M
@ 1999-09-30 18:02 ` Billinghurst, David (RTD)
1 sibling, 0 replies; 28+ messages in thread
From: Billinghurst, David (RTD) @ 1999-09-30 18:02 UTC (permalink / raw)
To: 'Lack Mr G M', N8TM; +Cc: gcc, beyerj
The default behaviour of my R10k based irix6.5 systems is defined in
/etc/compiler.defaults
to be -DEFAULT:abi=n32:isa=mips4:proc=r10k
You can generate -mips3 -n32 compiler and binaries by:
- changing /etc/compiler.defaults
- using the following patch to set the gcc default to -mips3
--- gcc/config/mips/iris6.h.0 Thu Jun 24 15:50:31 1999
+++ gcc/config/mips/iris6.h Thu Jun 24 15:56:20 1999
@@ -214,8 +214,9 @@
/* ??? If no mabi=X option give, but a mipsX option is, then should depend
on the mipsX option. */
+/* If no mips[3,4] option given, give the appropriate default for mabi=X */
#undef SUBTARGET_ASM_SPEC
-#define SUBTARGET_ASM_SPEC "%{!mabi*:-n32}"
+#define SUBTARGET_ASM_SPEC "%{!mabi*:-n32} %{!mips*: %{!mabi*:-mips3}
%{mabi=n32:-mips3} %{mabi=64:-mips4}}"
/* Must pass -g0 to the assembler, otherwise it may overwrite our
debug info with its own debug info. */
> -----Original Message-----
> From: Lack Mr G M [SMTP:gml4410@ggr.co.uk]
> Sent: Thursday, 2 September 1999 21:34
> To: N8TM@AOL.COM
> Cc: gcc@gcc.gnu.org; beyerj@sgi.com
> Subject: Re: Unsuccessful build of gcc-2.95.1
>
> N> > This is for mips-sgi-irix6.5 (claimed successful).
> > >
> > > The problem is in the configure stage.
> > >
> > > The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64
> ABI
> > > problems).
> > The lazy way is to set the environment variable CC='cc -n32' while using
> > MipsPro cc as a bootstrap compiler. AFAIK this works for MipsPro
> versions
> > since 7.2. Perhaps it's time to change a default configuration which
> caters
> > for the time before there was a working MipsPro compiler for n32/64.
> The
> > shoe probably is on the other foot now and 'cc -o32' may be less
> reliable.
>
> The stage1/xgcc is producing N32 mips-4 output, which fails to work
> when using the "cc -32" option in the mh-irix6 stub. Mind you, I've
> just noticed you suggest using "cc -n32"
>
> But I would still have a problem! I have networked systems (this is
> what Unix is all about..) and I need to build something which can run on
> any Irix6.5 system (it is NFS mounted to all of them). So mips-4
> binaries are out. How do I get gcc to produce -o32/mips-2 and
> -n32/mips-3 output? I have to link against libraries supplied by others
> (eg: Oracle), so can't expect n32 everywhere.
>
> James Beyer wrote:
> >
> > Can you send a "uname -a" for your machine?
>
> IRIX64 ukwsv100 6.5 04151556 IP27
>
> It's an O2000 with 32 R12k's
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-03 2:29 ` Lack Mr G M
@ 1999-09-30 18:02 ` Lack Mr G M
0 siblings, 0 replies; 28+ messages in thread
From: Lack Mr G M @ 1999-09-30 18:02 UTC (permalink / raw)
To: Billinghurst, David (RTD); +Cc: N8TM, gcc, beyerj
> The default behaviour of my R10k based irix6.5 systems is defined in
> /etc/compiler.defaults
> to be -DEFAULT:abi=n32:isa=mips4:proc=r10k
Yes. IMHO this is a mistake by SGI. The *default* should be
something that will run on most(all?) systems, so this should be
n32/mips/r4k. How many people will be shipping unusable binaries
because they don't know about this?
> You can generate -mips3 -n32 compiler and binaries by:
> - changing /etc/compiler.defaults
This assumes I have the rights to change the file.... The Makefile
should be able to pass on the compilation options I have asked it to
use.
> - using the following patch to set the gcc default to -mips3
Thanks. I'll give it a go (and try to work out what it is doing!).
PS: Building on a Challenge XL did work.
Also, the Solaris *build* works once I create the automount point
that the *install* will require. (Still can see no reson why the build
should be trying to use it).
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 10:00 ` James Beyer
@ 1999-09-30 18:02 ` James Beyer
0 siblings, 0 replies; 28+ messages in thread
From: James Beyer @ 1999-09-30 18:02 UTC (permalink / raw)
To: Lack Mr G M; +Cc: N8TM, gcc
I tried to push similar things through a little today and keep getting all
kinds of errors also.
The one work around that I used before was to use a Challenge system to
build gcc. This gave me an installation that would run on everything I
tried it on. So if all else fails and you have one try using a Challenge
system as your build system. Of course an INDY will work also but I tend
to shy away from building any large packages on an INDY anymore.
Once the build is complete you can run the "make install" on the O2000 and
everything "should" work. It did for me :;
Of course the prefered option would be getting configure and the Makefiles
working correctly. Something I am glad to test but I don't think I am the
best one to try mucking around in there right now.
good luck!
On Thu, 2 Sep 1999, Lack Mr G M wrote:
> Lack Mr G M wrote:
> >
> > Ok, using cc -n32 lets it run through, but it ends up producing a
> > mips4 binary, which isn't much use on an Indy, or a Challenge-XL. And
> > I;m of the opinion that I shoudl be able to have one installation, and
> > tell it what I want.
> >
> > So, I tried adding -mips3 to CFLAGS,
>
> Have now tried
>
> CC="cc -n32"
> CFLAGS="-O2 -mips3"
> CXXFLAGS="-O2 -mips3"
> GCC_CFLAGS="-mips3 -mabi=n32"
>
> as environment variables at both configrue and "make bootstrap" phases.
> It refuses to build anything other than a mips-4 crtbegin.o. (Could be
> because GCC_FLAGS isn't passed on between makes).
>
> So, I can't build a usable (for me) compiler.
>
> Mind you, I still have no idea why building on Solaris2.6 failed when
> it couldn't find an *installation* directory. Am I really the only
> person in the world who installs things under automount points? Mind
> you, why the -B options given to ./xgcc include these install
> directories in the first place is a bit odd. It is generated by these
> lines in gcc/Makefile:
>
> # Since tooldir does not exist at build-time, use
> -B$(build_tooldir)/bin/
> build_tooldir = $(exec_prefix)/$(target_alias)
>
> but this means that build_tooldir doesn't exist at build-time either,
> so no gain?
>
> OK: Just dicovered this is a bug which affects the Solaris build.
>
> If a directory doesn't exist in an automount point on Solaris you get
> "Permission denied", rather than the "No such file or directory" from
> Irix (and HP-UX). Presumably the code in xgcc does not allow for this
> (but why the Makefile should be using such a non-existent path at all
> with ./xgcc still remaina a mystery).
>
> Irix6.5:
> ukwit01*[1] ls /usr/central/non-existent-automountpoint
> Cannot access /usr/central/non-existent-automountpoint: No such file or
> directory
>
> Solaris2.6:
> uk0x05*[1] ls /usr/central/non-existent-automountpoint
> /usr/central/non-existent-automountpoint: Permission denied
>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 4:33 ` Lack Mr G M
1999-09-02 6:58 ` Lack Mr G M
@ 1999-09-30 18:02 ` Lack Mr G M
1 sibling, 0 replies; 28+ messages in thread
From: Lack Mr G M @ 1999-09-30 18:02 UTC (permalink / raw)
To: N8TM; +Cc: gcc, beyerj
N> > This is for mips-sgi-irix6.5 (claimed successful).
> >
> > The problem is in the configure stage.
> >
> > The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64 ABI
> > problems).
> The lazy way is to set the environment variable CC='cc -n32' while using
> MipsPro cc as a bootstrap compiler. AFAIK this works for MipsPro versions
> since 7.2. Perhaps it's time to change a default configuration which caters
> for the time before there was a working MipsPro compiler for n32/64. The
> shoe probably is on the other foot now and 'cc -o32' may be less reliable.
The stage1/xgcc is producing N32 mips-4 output, which fails to work
when using the "cc -32" option in the mh-irix6 stub. Mind you, I've
just noticed you suggest using "cc -n32"
But I would still have a problem! I have networked systems (this is
what Unix is all about..) and I need to build something which can run on
any Irix6.5 system (it is NFS mounted to all of them). So mips-4
binaries are out. How do I get gcc to produce -o32/mips-2 and
-n32/mips-3 output? I have to link against libraries supplied by others
(eg: Oracle), so can't expect n32 everywhere.
James Beyer wrote:
>
> Can you send a "uname -a" for your machine?
IRIX64 ukwsv100 6.5 04151556 IP27
It's an O2000 with 32 R12k's
^ permalink raw reply [flat|nested] 28+ messages in thread
* RE: Unsuccessful build of gcc-2.95.1
1999-09-01 20:41 Billinghurst, David (RTD)
1999-09-03 14:29 ` Erik Mouw
@ 1999-09-30 18:02 ` Billinghurst, David (RTD)
1 sibling, 0 replies; 28+ messages in thread
From: Billinghurst, David (RTD) @ 1999-09-30 18:02 UTC (permalink / raw)
To: 'gcc@gcc.gnu.org'
I can confirm this for MipsPro cc 7.2 and 7.2.1
> -----Original Message-----
> From: N8TM@aol.com [SMTP:N8TM@aol.com]
> Sent: Thursday, 2 September 1999 10:57
> To: gml4410@ggr.co.uk; gcc@gcc.gnu.org
> Subject: Re: Unsuccessful build of gcc-2.95.1
>
> In a message dated 9/1/99 1:16:54 PM EST, gml4410@ggr.co.uk writes:
>
> > This is for mips-sgi-irix6.5 (claimed successful).
> >
> > The problem is in the configure stage.
> >
> > The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64
> ABI
> > problems).
> The lazy way is to set the environment variable CC='cc -n32' while using
> MipsPro cc as a bootstrap compiler. AFAIK this works for MipsPro versions
>
> since 7.2. Perhaps it's time to change a default configuration which
> caters
> for the time before there was a working MipsPro compiler for n32/64. The
> shoe probably is on the other foot now and 'cc -o32' may be less reliable.
>
> Tim
> tprince@computer.org
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 6:58 ` Lack Mr G M
1999-09-02 7:09 ` James Beyer
1999-09-02 9:41 ` Lack Mr G M
@ 1999-09-30 18:02 ` Lack Mr G M
2 siblings, 0 replies; 28+ messages in thread
From: Lack Mr G M @ 1999-09-30 18:02 UTC (permalink / raw)
To: N8TM, gcc, beyerj
> The stage1/xgcc is producing N32 mips-4 output, which fails to work
> when using the "cc -32" option in the mh-irix6 stub. Mind you, I've
> just noticed you suggest using "cc -n32"
Ok, using cc -n32 lets it run through, but it ends up producing a
mips4 binary, which isn't much use on an Indy, or a Challenge-XL. And
I;m of the opinion that I shoudl be able to have one installation, and
tell it what I want.
So, I tried adding -mips3 to CFLAGS, That produces some warnign that
indicate it hasn't worked at stage1:
>> ld32: WARNING 158: Expecting MIPS3 objects: stage1/crtbegin.o is MIPS4.
some warnings which didn't occur without the -mips3
>> /usr/include/string.h:67: warning: conflicting types for
>> built-in function `memcpy'
and then the build core dumped with a Bus error.
>> ukwsv100*[1] file core
>> core: IRIX N32 core dump of 'gengenrtl'
The compilation options for this are attached.
stage1/xgcc -Bstage1/ -B/usr/central/mips-sgi-irix6.5/bin/ -DIN_GCC -O2 -O2 -mips3 -DHAVE_CONFIG_H -I. -I../../gcc-2.95.1/gcc -I../../gcc-2.95.1/gcc/config -I../../gcc-2.95.1/gcc/../include \
-DGCC_INCLUDE_DIR=\"/usr/central/lib/gcc-lib/mips-sgi-irix6.5/2.95.1/include\" \
-DGPLUSPLUS_INCLUDE_DIR=\"/usr/central/lib/gcc-lib/mips-sgi-irix6.5/2.95.1/../../../..`echo /usr/central | sed -e 's|^/usr/central||' -e 's|/[^/]*|/..|g'`/include/g++-3\" \
-DLOCAL_INCLUDE_DIR=\"/usr/local/include\" \
-DCROSS_INCLUDE_DIR=\"/usr/central/lib/gcc-lib/mips-sgi-irix6.5/2.95.1/../../../../mips-sgi-irix6.5/sys-include\" \
-DTOOL_INCLUDE_DIR=\"/usr/central/lib/gcc-lib/mips-sgi-irix6.5/2.95.1/../../../../mips-sgi-irix6.5/include\" \
-c `echo ../../gcc-2.95.1/gcc/cppinit.c | sed 's,^\./,,'`
../../gcc-2.95.1/gcc/cccp.c: In function `pcfinclude':
../../gcc-2.95.1/gcc/cccp.c:5485: warning: cast from pointer to integer of different size
../../gcc-2.95.1/gcc/cccp.c:5486: warning: cast from pointer to integer of different size
stage1/xgcc -Bstage1/ -B/usr/central/mips-sgi-irix6.5/bin/ -c -DIN_GCC -O2 -O2 -mips3 -DHAVE_CONFIG_H -I. -I../../gcc-2.95.1/gcc -I../../gcc-2.95.1/gcc/config -I../../gcc-2.95.1/gcc/../include ../../gcc-2.95.1/gcc/cppulp.c
stage1/xgcc -Bstage1/ -B/usr/central/mips-sgi-irix6.5/bin/ -c -DIN_GCC -O2 -O2 -mips3 -DHAVE_CONFIG_H -I. -I../../gcc-2.95.1/gcc -I../../gcc-2.95.1/gcc/config -I../../gcc-2.95.1/gcc/../include ../../gcc-2.95.1/gcc/scan-decls.c
./gencheck > tmp-check.h
./gengenrtl tmp-genrtl.h tmp-genrtl.c
make[2]: *** [s-genrtl] Bus error (core dumped)
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-01 11:55 ` James Beyer
@ 1999-09-30 18:02 ` James Beyer
0 siblings, 0 replies; 28+ messages in thread
From: James Beyer @ 1999-09-30 18:02 UTC (permalink / raw)
To: Lack Mr G M; +Cc: gcc
Can you send a "uname -a" for your machine? I have build gcc-2.95.1 on
several mips-sgi-irix6.5 machines without problems, during the build or
configure.
james
On Wed, 1 Sep 1999, Lack Mr G M wrote:
> This is for mips-sgi-irix6.5 (claimed successful).
>
> The problem is in the configure stage.
>
> The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64 ABI
> problems). However, after this is edited into the Makefile (sed -e
> /^####/ r mh-frag Makefile) another sed command runs an, at line 1382 of
> configure, this resets all CC definitions!
>
> I can add it back by hand, but....
>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 9:41 ` Lack Mr G M
1999-09-02 10:00 ` James Beyer
@ 1999-09-30 18:02 ` Lack Mr G M
1 sibling, 0 replies; 28+ messages in thread
From: Lack Mr G M @ 1999-09-30 18:02 UTC (permalink / raw)
To: N8TM, gcc, beyerj
Lack Mr G M wrote:
>
> Ok, using cc -n32 lets it run through, but it ends up producing a
> mips4 binary, which isn't much use on an Indy, or a Challenge-XL. And
> I;m of the opinion that I shoudl be able to have one installation, and
> tell it what I want.
>
> So, I tried adding -mips3 to CFLAGS,
Have now tried
CC="cc -n32"
CFLAGS="-O2 -mips3"
CXXFLAGS="-O2 -mips3"
GCC_CFLAGS="-mips3 -mabi=n32"
as environment variables at both configrue and "make bootstrap" phases.
It refuses to build anything other than a mips-4 crtbegin.o. (Could be
because GCC_FLAGS isn't passed on between makes).
So, I can't build a usable (for me) compiler.
Mind you, I still have no idea why building on Solaris2.6 failed when
it couldn't find an *installation* directory. Am I really the only
person in the world who installs things under automount points? Mind
you, why the -B options given to ./xgcc include these install
directories in the first place is a bit odd. It is generated by these
lines in gcc/Makefile:
# Since tooldir does not exist at build-time, use
-B$(build_tooldir)/bin/
build_tooldir = $(exec_prefix)/$(target_alias)
but this means that build_tooldir doesn't exist at build-time either,
so no gain?
OK: Just dicovered this is a bug which affects the Solaris build.
If a directory doesn't exist in an automount point on Solaris you get
"Permission denied", rather than the "No such file or directory" from
Irix (and HP-UX). Presumably the code in xgcc does not allow for this
(but why the Makefile should be using such a non-existent path at all
with ./xgcc still remaina a mystery).
Irix6.5:
ukwit01*[1] ls /usr/central/non-existent-automountpoint
Cannot access /usr/central/non-existent-automountpoint: No such file or
directory
Solaris2.6:
uk0x05*[1] ls /usr/central/non-existent-automountpoint
/usr/central/non-existent-automountpoint: Permission denied
^ permalink raw reply [flat|nested] 28+ messages in thread
* Unsuccessful build of gcc-2.95.1
1999-09-01 10:50 Lack Mr G M
1999-09-01 11:55 ` James Beyer
@ 1999-09-30 18:02 ` Lack Mr G M
1 sibling, 0 replies; 28+ messages in thread
From: Lack Mr G M @ 1999-09-30 18:02 UTC (permalink / raw)
To: gcc
This is for mips-sgi-irix6.5 (claimed successful).
The problem is in the configure stage.
The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64 ABI
problems). However, after this is edited into the Makefile (sed -e
/^####/ r mh-frag Makefile) another sed command runs an, at line 1382 of
configure, this resets all CC definitions!
I can add it back by hand, but....
^ permalink raw reply [flat|nested] 28+ messages in thread
* RE: Unsuccessful build of gcc-2.95.1
1999-09-03 14:29 ` Erik Mouw
@ 1999-09-30 18:02 ` Erik Mouw
0 siblings, 0 replies; 28+ messages in thread
From: Erik Mouw @ 1999-09-30 18:02 UTC (permalink / raw)
To: David.Billinghurst; +Cc: gcc
On Thu, 02 Sep 1999 03:09:26 +0000, RTD wrote:
> I can confirm this for MipsPro cc 7.2 and 7.2.1
I can't. A build on an Octane MXI running IRIX 6.5.3 using MipsPRO C 7.2.1
compiled a clean (although not yet heavily tested) gcc-2.95.1. AFAIK every
MipsPRO compiler comes with an o32 backend.
Erik
>> -----Original Message-----
>> From: N8TM@aol.com [SMTP:N8TM@aol.com]
>> Sent: Thursday, 2 September 1999 10:57
>> To: gml4410@ggr.co.uk; gcc@gcc.gnu.org
>> Subject: Re: Unsuccessful build of gcc-2.95.1
>>
>> In a message dated 9/1/99 1:16:54 PM EST, gml4410@ggr.co.uk writes:
>>
>> > This is for mips-sgi-irix6.5 (claimed successful).
>> >
>> > The problem is in the configure stage.
>> >
>> > The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64
>> ABI
>> > problems).
>> The lazy way is to set the environment variable CC='cc -n32' while using
>> MipsPro cc as a bootstrap compiler. AFAIK this works for MipsPro versions
>>
>> since 7.2. Perhaps it's time to change a default configuration which
>> caters
>> for the time before there was a working MipsPro compiler for n32/64. The
>> shoe probably is on the other foot now and 'cc -o32' may be less reliable.
>>
>> Tim
>> tprince@computer.org
>
--
J.A.K. (Erik) Mouw, Information and Communication Theory Group, Department
of Electrical Engineering, Faculty of Information Technology and Systems,
Delft University of Technology, PO BOX 5031, 2600 GA Delft, The Netherlands
Phone: +31-15-2785859 Fax: +31-15-2781843 Email J.A.K.Mouw@its.tudelft.nl
WWW: http://www-ict.its.tudelft.nl/~erik/
^ permalink raw reply [flat|nested] 28+ messages in thread
* RE: Unsuccessful build of gcc-2.95.1
1999-09-01 20:41 Billinghurst, David (RTD)
@ 1999-09-03 14:29 ` Erik Mouw
1999-09-30 18:02 ` Erik Mouw
1999-09-30 18:02 ` Billinghurst, David (RTD)
1 sibling, 1 reply; 28+ messages in thread
From: Erik Mouw @ 1999-09-03 14:29 UTC (permalink / raw)
To: David.Billinghurst; +Cc: gcc
On Thu, 02 Sep 1999 03:09:26 +0000, RTD wrote:
> I can confirm this for MipsPro cc 7.2 and 7.2.1
I can't. A build on an Octane MXI running IRIX 6.5.3 using MipsPRO C 7.2.1
compiled a clean (although not yet heavily tested) gcc-2.95.1. AFAIK every
MipsPRO compiler comes with an o32 backend.
Erik
>> -----Original Message-----
>> From: N8TM@aol.com [SMTP:N8TM@aol.com]
>> Sent: Thursday, 2 September 1999 10:57
>> To: gml4410@ggr.co.uk; gcc@gcc.gnu.org
>> Subject: Re: Unsuccessful build of gcc-2.95.1
>>
>> In a message dated 9/1/99 1:16:54 PM EST, gml4410@ggr.co.uk writes:
>>
>> > This is for mips-sgi-irix6.5 (claimed successful).
>> >
>> > The problem is in the configure stage.
>> >
>> > The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64
>> ABI
>> > problems).
>> The lazy way is to set the environment variable CC='cc -n32' while using
>> MipsPro cc as a bootstrap compiler. AFAIK this works for MipsPro versions
>>
>> since 7.2. Perhaps it's time to change a default configuration which
>> caters
>> for the time before there was a working MipsPro compiler for n32/64. The
>> shoe probably is on the other foot now and 'cc -o32' may be less reliable.
>>
>> Tim
>> tprince@computer.org
>
--
J.A.K. (Erik) Mouw, Information and Communication Theory Group, Department
of Electrical Engineering, Faculty of Information Technology and Systems,
Delft University of Technology, PO BOX 5031, 2600 GA Delft, The Netherlands
Phone: +31-15-2785859 Fax: +31-15-2781843 Email J.A.K.Mouw@its.tudelft.nl
WWW: http://www-ict.its.tudelft.nl/~erik/
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 16:10 Billinghurst, David (RTD)
@ 1999-09-03 2:29 ` Lack Mr G M
1999-09-30 18:02 ` Lack Mr G M
1999-09-30 18:02 ` Billinghurst, David (RTD)
1 sibling, 1 reply; 28+ messages in thread
From: Lack Mr G M @ 1999-09-03 2:29 UTC (permalink / raw)
To: Billinghurst, David (RTD); +Cc: N8TM, gcc, beyerj
> The default behaviour of my R10k based irix6.5 systems is defined in
> /etc/compiler.defaults
> to be -DEFAULT:abi=n32:isa=mips4:proc=r10k
Yes. IMHO this is a mistake by SGI. The *default* should be
something that will run on most(all?) systems, so this should be
n32/mips/r4k. How many people will be shipping unusable binaries
because they don't know about this?
> You can generate -mips3 -n32 compiler and binaries by:
> - changing /etc/compiler.defaults
This assumes I have the rights to change the file.... The Makefile
should be able to pass on the compilation options I have asked it to
use.
> - using the following patch to set the gcc default to -mips3
Thanks. I'll give it a go (and try to work out what it is doing!).
PS: Building on a Challenge XL did work.
Also, the Solaris *build* works once I create the automount point
that the *install* will require. (Still can see no reson why the build
should be trying to use it).
^ permalink raw reply [flat|nested] 28+ messages in thread
* RE: Unsuccessful build of gcc-2.95.1
@ 1999-09-02 16:10 Billinghurst, David (RTD)
1999-09-03 2:29 ` Lack Mr G M
1999-09-30 18:02 ` Billinghurst, David (RTD)
0 siblings, 2 replies; 28+ messages in thread
From: Billinghurst, David (RTD) @ 1999-09-02 16:10 UTC (permalink / raw)
To: 'Lack Mr G M', N8TM; +Cc: gcc, beyerj
The default behaviour of my R10k based irix6.5 systems is defined in
/etc/compiler.defaults
to be -DEFAULT:abi=n32:isa=mips4:proc=r10k
You can generate -mips3 -n32 compiler and binaries by:
- changing /etc/compiler.defaults
- using the following patch to set the gcc default to -mips3
--- gcc/config/mips/iris6.h.0 Thu Jun 24 15:50:31 1999
+++ gcc/config/mips/iris6.h Thu Jun 24 15:56:20 1999
@@ -214,8 +214,9 @@
/* ??? If no mabi=X option give, but a mipsX option is, then should depend
on the mipsX option. */
+/* If no mips[3,4] option given, give the appropriate default for mabi=X */
#undef SUBTARGET_ASM_SPEC
-#define SUBTARGET_ASM_SPEC "%{!mabi*:-n32}"
+#define SUBTARGET_ASM_SPEC "%{!mabi*:-n32} %{!mips*: %{!mabi*:-mips3}
%{mabi=n32:-mips3} %{mabi=64:-mips4}}"
/* Must pass -g0 to the assembler, otherwise it may overwrite our
debug info with its own debug info. */
> -----Original Message-----
> From: Lack Mr G M [SMTP:gml4410@ggr.co.uk]
> Sent: Thursday, 2 September 1999 21:34
> To: N8TM@AOL.COM
> Cc: gcc@gcc.gnu.org; beyerj@sgi.com
> Subject: Re: Unsuccessful build of gcc-2.95.1
>
> N> > This is for mips-sgi-irix6.5 (claimed successful).
> > >
> > > The problem is in the configure stage.
> > >
> > > The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64
> ABI
> > > problems).
> > The lazy way is to set the environment variable CC='cc -n32' while using
> > MipsPro cc as a bootstrap compiler. AFAIK this works for MipsPro
> versions
> > since 7.2. Perhaps it's time to change a default configuration which
> caters
> > for the time before there was a working MipsPro compiler for n32/64.
> The
> > shoe probably is on the other foot now and 'cc -o32' may be less
> reliable.
>
> The stage1/xgcc is producing N32 mips-4 output, which fails to work
> when using the "cc -32" option in the mh-irix6 stub. Mind you, I've
> just noticed you suggest using "cc -n32"
>
> But I would still have a problem! I have networked systems (this is
> what Unix is all about..) and I need to build something which can run on
> any Irix6.5 system (it is NFS mounted to all of them). So mips-4
> binaries are out. How do I get gcc to produce -o32/mips-2 and
> -n32/mips-3 output? I have to link against libraries supplied by others
> (eg: Oracle), so can't expect n32 everywhere.
>
> James Beyer wrote:
> >
> > Can you send a "uname -a" for your machine?
>
> IRIX64 ukwsv100 6.5 04151556 IP27
>
> It's an O2000 with 32 R12k's
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 9:41 ` Lack Mr G M
@ 1999-09-02 10:00 ` James Beyer
1999-09-30 18:02 ` James Beyer
1999-09-30 18:02 ` Lack Mr G M
1 sibling, 1 reply; 28+ messages in thread
From: James Beyer @ 1999-09-02 10:00 UTC (permalink / raw)
To: Lack Mr G M; +Cc: N8TM, gcc
I tried to push similar things through a little today and keep getting all
kinds of errors also.
The one work around that I used before was to use a Challenge system to
build gcc. This gave me an installation that would run on everything I
tried it on. So if all else fails and you have one try using a Challenge
system as your build system. Of course an INDY will work also but I tend
to shy away from building any large packages on an INDY anymore.
Once the build is complete you can run the "make install" on the O2000 and
everything "should" work. It did for me :;
Of course the prefered option would be getting configure and the Makefiles
working correctly. Something I am glad to test but I don't think I am the
best one to try mucking around in there right now.
good luck!
On Thu, 2 Sep 1999, Lack Mr G M wrote:
> Lack Mr G M wrote:
> >
> > Ok, using cc -n32 lets it run through, but it ends up producing a
> > mips4 binary, which isn't much use on an Indy, or a Challenge-XL. And
> > I;m of the opinion that I shoudl be able to have one installation, and
> > tell it what I want.
> >
> > So, I tried adding -mips3 to CFLAGS,
>
> Have now tried
>
> CC="cc -n32"
> CFLAGS="-O2 -mips3"
> CXXFLAGS="-O2 -mips3"
> GCC_CFLAGS="-mips3 -mabi=n32"
>
> as environment variables at both configrue and "make bootstrap" phases.
> It refuses to build anything other than a mips-4 crtbegin.o. (Could be
> because GCC_FLAGS isn't passed on between makes).
>
> So, I can't build a usable (for me) compiler.
>
> Mind you, I still have no idea why building on Solaris2.6 failed when
> it couldn't find an *installation* directory. Am I really the only
> person in the world who installs things under automount points? Mind
> you, why the -B options given to ./xgcc include these install
> directories in the first place is a bit odd. It is generated by these
> lines in gcc/Makefile:
>
> # Since tooldir does not exist at build-time, use
> -B$(build_tooldir)/bin/
> build_tooldir = $(exec_prefix)/$(target_alias)
>
> but this means that build_tooldir doesn't exist at build-time either,
> so no gain?
>
> OK: Just dicovered this is a bug which affects the Solaris build.
>
> If a directory doesn't exist in an automount point on Solaris you get
> "Permission denied", rather than the "No such file or directory" from
> Irix (and HP-UX). Presumably the code in xgcc does not allow for this
> (but why the Makefile should be using such a non-existent path at all
> with ./xgcc still remaina a mystery).
>
> Irix6.5:
> ukwit01*[1] ls /usr/central/non-existent-automountpoint
> Cannot access /usr/central/non-existent-automountpoint: No such file or
> directory
>
> Solaris2.6:
> uk0x05*[1] ls /usr/central/non-existent-automountpoint
> /usr/central/non-existent-automountpoint: Permission denied
>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 6:58 ` Lack Mr G M
1999-09-02 7:09 ` James Beyer
@ 1999-09-02 9:41 ` Lack Mr G M
1999-09-02 10:00 ` James Beyer
1999-09-30 18:02 ` Lack Mr G M
1999-09-30 18:02 ` Lack Mr G M
2 siblings, 2 replies; 28+ messages in thread
From: Lack Mr G M @ 1999-09-02 9:41 UTC (permalink / raw)
To: N8TM, gcc, beyerj
Lack Mr G M wrote:
>
> Ok, using cc -n32 lets it run through, but it ends up producing a
> mips4 binary, which isn't much use on an Indy, or a Challenge-XL. And
> I;m of the opinion that I shoudl be able to have one installation, and
> tell it what I want.
>
> So, I tried adding -mips3 to CFLAGS,
Have now tried
CC="cc -n32"
CFLAGS="-O2 -mips3"
CXXFLAGS="-O2 -mips3"
GCC_CFLAGS="-mips3 -mabi=n32"
as environment variables at both configrue and "make bootstrap" phases.
It refuses to build anything other than a mips-4 crtbegin.o. (Could be
because GCC_FLAGS isn't passed on between makes).
So, I can't build a usable (for me) compiler.
Mind you, I still have no idea why building on Solaris2.6 failed when
it couldn't find an *installation* directory. Am I really the only
person in the world who installs things under automount points? Mind
you, why the -B options given to ./xgcc include these install
directories in the first place is a bit odd. It is generated by these
lines in gcc/Makefile:
# Since tooldir does not exist at build-time, use
-B$(build_tooldir)/bin/
build_tooldir = $(exec_prefix)/$(target_alias)
but this means that build_tooldir doesn't exist at build-time either,
so no gain?
OK: Just dicovered this is a bug which affects the Solaris build.
If a directory doesn't exist in an automount point on Solaris you get
"Permission denied", rather than the "No such file or directory" from
Irix (and HP-UX). Presumably the code in xgcc does not allow for this
(but why the Makefile should be using such a non-existent path at all
with ./xgcc still remaina a mystery).
Irix6.5:
ukwit01*[1] ls /usr/central/non-existent-automountpoint
Cannot access /usr/central/non-existent-automountpoint: No such file or
directory
Solaris2.6:
uk0x05*[1] ls /usr/central/non-existent-automountpoint
/usr/central/non-existent-automountpoint: Permission denied
^ permalink raw reply [flat|nested] 28+ messages in thread
* Unsuccessful build of gcc-2.95.1
@ 1999-09-02 9:19 Anna Winkler
1999-09-30 18:02 ` Anna Winkler
0 siblings, 1 reply; 28+ messages in thread
From: Anna Winkler @ 1999-09-02 9:19 UTC (permalink / raw)
To: GCC List; +Cc: gml4410
> Date: Wed, 01 Sep 1999 19:40:43 +0100
> From: Lack Mr G M <gml4410@ggr.co.uk>
> To: gcc@gcc.gnu.org
> Subject: Unsuccessful build of gcc-2.95.1 (more of them).
>
> Is it just me???
>
> Having edited the Makefile to set "CC = cc -32" for mips-sgi-irix6.5
> the build bombs out with:
>
> > ld32: FATAL 12: Expecting n32 objects:
> > ../libiberty/libiberty.a(pexecute.o) is o32.
>
> so something is trying to use non-"-32" anyway...
You want to use -n32 instead of -32. -32 and -o32 mean generate an old
32-bit object, while -n32 is a new 32-bit object.
--Anna Winkler
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 6:58 ` Lack Mr G M
@ 1999-09-02 7:09 ` James Beyer
1999-09-30 18:02 ` James Beyer
1999-09-02 9:41 ` Lack Mr G M
1999-09-30 18:02 ` Lack Mr G M
2 siblings, 1 reply; 28+ messages in thread
From: James Beyer @ 1999-09-02 7:09 UTC (permalink / raw)
To: Lack Mr G M; +Cc: N8TM, gcc
On Thu, 2 Sep 1999, Lack Mr G M wrote:
> > The stage1/xgcc is producing N32 mips-4 output, which fails to work
> > when using the "cc -32" option in the mh-irix6 stub. Mind you, I've
> > just noticed you suggest using "cc -n32"
>
> Ok, using cc -n32 lets it run through, but it ends up producing a
> mips4 binary, which isn't much use on an Indy, or a Challenge-XL. And
> I;m of the opinion that I shoudl be able to have one installation, and
> tell it what I want.
>
> So, I tried adding -mips3 to CFLAGS, That produces some warnign that
> indicate it hasn't worked at stage1:
>
> >> ld32: WARNING 158: Expecting MIPS3 objects: stage1/crtbegin.o is MIPS4.
Try using cc -n32 -mips3. Just using -n32 on an ORIGIN will default to
-mips4. Whereas a the other generates code that runs on my INDY.
>
> some warnings which didn't occur without the -mips3
>
> >> /usr/include/string.h:67: warning: conflicting types for
> >> built-in function `memcpy'
>
This is because of gcc's built-ins, and is something I have learned to
deal with in alot of my code. If a -fno-builtin is given to gcc it will
go away. Although I would prefer a way to quelch the message and have gcc
use its builtin calls if they truely are better.
Anyone have a better answer here?
> and then the build core dumped with a Bus error.
>
> >> ukwsv100*[1] file core
> >> core: IRIX N32 core dump of 'gengenrtl'
>
> The compilation options for this are attached.
I hope someone else knows something here!
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-02 4:33 ` Lack Mr G M
@ 1999-09-02 6:58 ` Lack Mr G M
1999-09-02 7:09 ` James Beyer
` (2 more replies)
1999-09-30 18:02 ` Lack Mr G M
1 sibling, 3 replies; 28+ messages in thread
From: Lack Mr G M @ 1999-09-02 6:58 UTC (permalink / raw)
To: N8TM, gcc, beyerj
> The stage1/xgcc is producing N32 mips-4 output, which fails to work
> when using the "cc -32" option in the mh-irix6 stub. Mind you, I've
> just noticed you suggest using "cc -n32"
Ok, using cc -n32 lets it run through, but it ends up producing a
mips4 binary, which isn't much use on an Indy, or a Challenge-XL. And
I;m of the opinion that I shoudl be able to have one installation, and
tell it what I want.
So, I tried adding -mips3 to CFLAGS, That produces some warnign that
indicate it hasn't worked at stage1:
>> ld32: WARNING 158: Expecting MIPS3 objects: stage1/crtbegin.o is MIPS4.
some warnings which didn't occur without the -mips3
>> /usr/include/string.h:67: warning: conflicting types for
>> built-in function `memcpy'
and then the build core dumped with a Bus error.
>> ukwsv100*[1] file core
>> core: IRIX N32 core dump of 'gengenrtl'
The compilation options for this are attached.
stage1/xgcc -Bstage1/ -B/usr/central/mips-sgi-irix6.5/bin/ -DIN_GCC -O2 -O2 -mips3 -DHAVE_CONFIG_H -I. -I../../gcc-2.95.1/gcc -I../../gcc-2.95.1/gcc/config -I../../gcc-2.95.1/gcc/../include \
-DGCC_INCLUDE_DIR=\"/usr/central/lib/gcc-lib/mips-sgi-irix6.5/2.95.1/include\" \
-DGPLUSPLUS_INCLUDE_DIR=\"/usr/central/lib/gcc-lib/mips-sgi-irix6.5/2.95.1/../../../..`echo /usr/central | sed -e 's|^/usr/central||' -e 's|/[^/]*|/..|g'`/include/g++-3\" \
-DLOCAL_INCLUDE_DIR=\"/usr/local/include\" \
-DCROSS_INCLUDE_DIR=\"/usr/central/lib/gcc-lib/mips-sgi-irix6.5/2.95.1/../../../../mips-sgi-irix6.5/sys-include\" \
-DTOOL_INCLUDE_DIR=\"/usr/central/lib/gcc-lib/mips-sgi-irix6.5/2.95.1/../../../../mips-sgi-irix6.5/include\" \
-c `echo ../../gcc-2.95.1/gcc/cppinit.c | sed 's,^\./,,'`
../../gcc-2.95.1/gcc/cccp.c: In function `pcfinclude':
../../gcc-2.95.1/gcc/cccp.c:5485: warning: cast from pointer to integer of different size
../../gcc-2.95.1/gcc/cccp.c:5486: warning: cast from pointer to integer of different size
stage1/xgcc -Bstage1/ -B/usr/central/mips-sgi-irix6.5/bin/ -c -DIN_GCC -O2 -O2 -mips3 -DHAVE_CONFIG_H -I. -I../../gcc-2.95.1/gcc -I../../gcc-2.95.1/gcc/config -I../../gcc-2.95.1/gcc/../include ../../gcc-2.95.1/gcc/cppulp.c
stage1/xgcc -Bstage1/ -B/usr/central/mips-sgi-irix6.5/bin/ -c -DIN_GCC -O2 -O2 -mips3 -DHAVE_CONFIG_H -I. -I../../gcc-2.95.1/gcc -I../../gcc-2.95.1/gcc/config -I../../gcc-2.95.1/gcc/../include ../../gcc-2.95.1/gcc/scan-decls.c
./gencheck > tmp-check.h
./gengenrtl tmp-genrtl.h tmp-genrtl.c
make[2]: *** [s-genrtl] Bus error (core dumped)
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-01 17:57 N8TM
@ 1999-09-02 4:33 ` Lack Mr G M
1999-09-02 6:58 ` Lack Mr G M
1999-09-30 18:02 ` Lack Mr G M
1999-09-30 18:02 ` N8TM
1 sibling, 2 replies; 28+ messages in thread
From: Lack Mr G M @ 1999-09-02 4:33 UTC (permalink / raw)
To: N8TM; +Cc: gcc, beyerj
N> > This is for mips-sgi-irix6.5 (claimed successful).
> >
> > The problem is in the configure stage.
> >
> > The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64 ABI
> > problems).
> The lazy way is to set the environment variable CC='cc -n32' while using
> MipsPro cc as a bootstrap compiler. AFAIK this works for MipsPro versions
> since 7.2. Perhaps it's time to change a default configuration which caters
> for the time before there was a working MipsPro compiler for n32/64. The
> shoe probably is on the other foot now and 'cc -o32' may be less reliable.
The stage1/xgcc is producing N32 mips-4 output, which fails to work
when using the "cc -32" option in the mh-irix6 stub. Mind you, I've
just noticed you suggest using "cc -n32"
But I would still have a problem! I have networked systems (this is
what Unix is all about..) and I need to build something which can run on
any Irix6.5 system (it is NFS mounted to all of them). So mips-4
binaries are out. How do I get gcc to produce -o32/mips-2 and
-n32/mips-3 output? I have to link against libraries supplied by others
(eg: Oracle), so can't expect n32 everywhere.
James Beyer wrote:
>
> Can you send a "uname -a" for your machine?
IRIX64 ukwsv100 6.5 04151556 IP27
It's an O2000 with 32 R12k's
^ permalink raw reply [flat|nested] 28+ messages in thread
* RE: Unsuccessful build of gcc-2.95.1
@ 1999-09-01 20:41 Billinghurst, David (RTD)
1999-09-03 14:29 ` Erik Mouw
1999-09-30 18:02 ` Billinghurst, David (RTD)
0 siblings, 2 replies; 28+ messages in thread
From: Billinghurst, David (RTD) @ 1999-09-01 20:41 UTC (permalink / raw)
To: 'gcc@gcc.gnu.org'
I can confirm this for MipsPro cc 7.2 and 7.2.1
> -----Original Message-----
> From: N8TM@aol.com [SMTP:N8TM@aol.com]
> Sent: Thursday, 2 September 1999 10:57
> To: gml4410@ggr.co.uk; gcc@gcc.gnu.org
> Subject: Re: Unsuccessful build of gcc-2.95.1
>
> In a message dated 9/1/99 1:16:54 PM EST, gml4410@ggr.co.uk writes:
>
> > This is for mips-sgi-irix6.5 (claimed successful).
> >
> > The problem is in the configure stage.
> >
> > The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64
> ABI
> > problems).
> The lazy way is to set the environment variable CC='cc -n32' while using
> MipsPro cc as a bootstrap compiler. AFAIK this works for MipsPro versions
>
> since 7.2. Perhaps it's time to change a default configuration which
> caters
> for the time before there was a working MipsPro compiler for n32/64. The
> shoe probably is on the other foot now and 'cc -o32' may be less reliable.
>
> Tim
> tprince@computer.org
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
@ 1999-09-01 17:57 N8TM
1999-09-02 4:33 ` Lack Mr G M
1999-09-30 18:02 ` N8TM
0 siblings, 2 replies; 28+ messages in thread
From: N8TM @ 1999-09-01 17:57 UTC (permalink / raw)
To: gml4410, gcc
In a message dated 9/1/99 1:16:54 PM EST, gml4410@ggr.co.uk writes:
> This is for mips-sgi-irix6.5 (claimed successful).
>
> The problem is in the configure stage.
>
> The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64 ABI
> problems).
The lazy way is to set the environment variable CC='cc -n32' while using
MipsPro cc as a bootstrap compiler. AFAIK this works for MipsPro versions
since 7.2. Perhaps it's time to change a default configuration which caters
for the time before there was a working MipsPro compiler for n32/64. The
shoe probably is on the other foot now and 'cc -o32' may be less reliable.
Tim
tprince@computer.org
^ permalink raw reply [flat|nested] 28+ messages in thread
* Re: Unsuccessful build of gcc-2.95.1
1999-09-01 10:50 Lack Mr G M
@ 1999-09-01 11:55 ` James Beyer
1999-09-30 18:02 ` James Beyer
1999-09-30 18:02 ` Lack Mr G M
1 sibling, 1 reply; 28+ messages in thread
From: James Beyer @ 1999-09-01 11:55 UTC (permalink / raw)
To: Lack Mr G M; +Cc: gcc
Can you send a "uname -a" for your machine? I have build gcc-2.95.1 on
several mips-sgi-irix6.5 machines without problems, during the build or
configure.
james
On Wed, 1 Sep 1999, Lack Mr G M wrote:
> This is for mips-sgi-irix6.5 (claimed successful).
>
> The problem is in the configure stage.
>
> The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64 ABI
> problems). However, after this is edited into the Makefile (sed -e
> /^####/ r mh-frag Makefile) another sed command runs an, at line 1382 of
> configure, this resets all CC definitions!
>
> I can add it back by hand, but....
>
^ permalink raw reply [flat|nested] 28+ messages in thread
* Unsuccessful build of gcc-2.95.1
@ 1999-09-01 10:50 Lack Mr G M
1999-09-01 11:55 ` James Beyer
1999-09-30 18:02 ` Lack Mr G M
0 siblings, 2 replies; 28+ messages in thread
From: Lack Mr G M @ 1999-09-01 10:50 UTC (permalink / raw)
To: gcc
This is for mips-sgi-irix6.5 (claimed successful).
The problem is in the configure stage.
The config/mh-irix6 tries to set "CC = cc -32" (to avoid n32/n64 ABI
problems). However, after this is edited into the Makefile (sed -e
/^####/ r mh-frag Makefile) another sed command runs an, at line 1382 of
configure, this resets all CC definitions!
I can add it back by hand, but....
^ permalink raw reply [flat|nested] 28+ messages in thread
end of thread, other threads:[~1999-09-30 18:02 UTC | newest]
Thread overview: 28+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
1999-09-02 19:35 Unsuccessful build of gcc-2.95.1 N8TM
1999-09-30 18:02 ` N8TM
-- strict thread matches above, loose matches on Subject: below --
1999-09-02 16:10 Billinghurst, David (RTD)
1999-09-03 2:29 ` Lack Mr G M
1999-09-30 18:02 ` Lack Mr G M
1999-09-30 18:02 ` Billinghurst, David (RTD)
1999-09-02 9:19 Anna Winkler
1999-09-30 18:02 ` Anna Winkler
1999-09-01 20:41 Billinghurst, David (RTD)
1999-09-03 14:29 ` Erik Mouw
1999-09-30 18:02 ` Erik Mouw
1999-09-30 18:02 ` Billinghurst, David (RTD)
1999-09-01 17:57 N8TM
1999-09-02 4:33 ` Lack Mr G M
1999-09-02 6:58 ` Lack Mr G M
1999-09-02 7:09 ` James Beyer
1999-09-30 18:02 ` James Beyer
1999-09-02 9:41 ` Lack Mr G M
1999-09-02 10:00 ` James Beyer
1999-09-30 18:02 ` James Beyer
1999-09-30 18:02 ` Lack Mr G M
1999-09-30 18:02 ` Lack Mr G M
1999-09-30 18:02 ` Lack Mr G M
1999-09-30 18:02 ` N8TM
1999-09-01 10:50 Lack Mr G M
1999-09-01 11:55 ` James Beyer
1999-09-30 18:02 ` James Beyer
1999-09-30 18:02 ` Lack Mr G M
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).