back to gcc under pre-R5

back to gcc under pre-R5

[Andreas Tobler]

Hello,

is it possible to get back to gcc under pre-R5? If so. where can I get the (S)RPMS?

And how much do it affect when I build the libstdc++ with fno-vtable-thunks?
Do I have to rebuild the whole system, since the new glibc needs vtables?
It's only because egcs has bug which is for my subject killing.

Thanks for any hint

Andreas
-- 
------------------
| Andreas Tobler								
| CH-8004 Zuerich	
| E-mail: toa@pop.agri.ch		
------------------------------------------

[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to  Cc linuxppc-dev  if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]

RE: back to gcc under pre-R5

[Gary Thomas]

On 17-Feb-99 Andreas Tobler wrote:
> 
> Hello,
> 
> is it possible to get back to gcc under pre-R5? If so. where can I get the (S)RPMS?
> 
> And how much do it affect when I build the libstdc++ with fno-vtable-thunks?
> Do I have to rebuild the whole system, since the new glibc needs vtables?
> It's only because egcs has bug which is for my subject killing.
> 

There are no current RPMs for GCC to my knowledge.  The last ones I made nearly
two years ago and gave up on shortly thereafter because EGCS was much superior.

What bug in EGCS is so terrible that you are considering this move?
Which version of EGCS have you tried?

There will be EGCS-1.1.2 in the very near future which has many PowerPC problems
fixed, perhaps you can wait.

------------------------------------------------------------------------
Gary Thomas                              |
email: gdt@linuxppc.org                  | "Fine wine is a necessity of
   ... opinions expressed here are mine  |        life for me"
       and no one else would claim them! |
                                         |      Thomas Jefferson
------------------------------------------------------------------------



[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to  Cc linuxppc-dev  if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]

Re: back to gcc under pre-R5

[Andreas Tobler]

Gary Thomas wrote:
> 
> On 17-Feb-99 Andreas Tobler wrote:
> >
> > Hello,
> >
> > is it possible to get back to gcc under pre-R5? If so. where can I get the (S)RPMS?
> >
> > And how much do it affect when I build the libstdc++ with fno-vtable-thunks?
> > Do I have to rebuild the whole system, since the new glibc needs vtables?
> > It's only because egcs has bug which is for my subject killing.
> >
> 
> There are no current RPMs for GCC to my knowledge.  The last ones I made nearly
> two years ago and gave up on shortly thereafter because EGCS was much superior.
> 

I found a rpm with gcc-2.8.0-1.src.rpm, maybe??
I think it was on Linuxppc mirror in uk in the 'linuxppc-delete-me' folder.

> What bug in EGCS is so terrible that you are considering this move?
> Which version of EGCS have you tried?

It's not quite a bug, it's an optimization which is not implemented yet. The
vtable syndrom!
I tried egcs-1.0.3 under R4 there it works.
Also tried egcs-1.1.1.c on pre-R5 no success.
Built egcs -1.1.1ca by myself with vtable thunks off, doesn't work because
glibc needs vtable thunks. (Interpretation from my side when Franz Sirl
suggested to rebuild the glibc) 

The whole story happens when I build JX (http://www.cco.caltech.edu/~jafl/jx/jx2.html)

CC get's a core dump in a few points. 

I'd like to build this Code Crusader out of personal interest, maybe we can
share my experiences with others when they find to JX? 
> 
> There will be EGCS-1.1.2 in the very near future which has many PowerPC problems
> fixed, perhaps you can wait.

I could wait if this thing would be fixed, but as Mike Stump said, it wouldn't
be fixed in the near future.

Thanks
Andreas
-- 
------------------
| Andreas Tobler								
| CH-8004 Zuerich	
| E-mail: toa@pop.agri.ch		
------------------------------------------

[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to  Cc linuxppc-dev  if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]

bugs building stuff.. somewhere.. (was: back to gcc under pre-R5)

[Brian Downing]

On Wed, 17 Feb 1999 13:34:31 -0000 (GMT), Gary Thomas writes:
 > There are no current RPMs for GCC to my knowledge.  The last ones I made
 > nearly two years ago and gave up on shortly thereafter because EGCS was
 > much superior.
 > 
 > What bug in EGCS is so terrible that you are considering this move?
 > Which version of EGCS have you tried?
 > 
 > There will be EGCS-1.1.2 in the very near future which has many PowerPC
 > problems fixed, perhaps you can wait.

I've had similar problems building software.  I'm not sure where the
problem is, but I cannot compile this software statically.  I wouldn't
care, but for it to be useful, it must be able to run on all PPC
platforms.

When I link dynamically, everything works fine, everywhere that I've
built it - unfortunately binary portability is a little low.

When I try to link it statically, I get a bunch of ld errors from
libc.a about "tried to relocate __pthread_mutex_lock" and other various
__pthread_* calls.  No binary is generated.  I tried to compile a
non-reentrant version of glibc with no linuxthreads (since I'm building
statically and the software isn't threaded I figured it was worth a try).
Unfortunately glibc doesn't appear to build without linuxthreads anymore
- even with --disable-sanity-checks.  I've even gone so far as to ifdef
out the offending calls in glibc (for testing), and other similar errors
show up for different places in glibc.  I don't have the exact errors
on hand right now, but I can get them if anyone thinks they would help.

Of course, I've tried compiling on MkLinux, and it builds fine, but a
different component dumps core when built there.  Of course, the same
component when compiled statically on LinuxPPC-pre-r5, will run on
MkLinux, but NOT LinuxPPC-r4 (Illegal Instruction).

So maybe I should ask the evil question:  Just how binary compatible
are the various PPC kernels and distributions.

(Built on a PowerBook G3 266, with linuxppc-pre-r5, with egcs-1.1.1 and
the latest snapshot I could build, with glibc-2.0.108 and glibc-2.1,
and with some other variations made for the heck of it.)

So anyway, sorry for the sob story, but in the last two weeks I've seen
more failed builds for dumb reasons... :)

-bcd
--
**               Brian Downing
**  UNIX Systems Administrator
**        bdowning@wolfram.com

[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to  Cc linuxppc-dev  if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]

Re: back to gcc under pre-R5

[Tom Rini]

On Wed, 17 Feb 1999, Andreas Tobler wrote:

> is it possible to get back to gcc under pre-R5? If so. where can I get the (S)RPMS?

Erm, gcc 2.7.2.3?  You could, but that's more buggy then egcs. :)

> And how much do it affect when I build the libstdc++ with fno-vtable-thunks?
> Do I have to rebuild the whole system, since the new glibc needs vtables?
> It's only because egcs has bug which is for my subject killing.

What bug are you running into?  It's better to get 'em fixed now then try
to work around 'em...

---
Tom Rini (TR1265)
http://dobbstown.yeti.edu/


[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to  Cc linuxppc-dev  if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]

Re: back to gcc under pre-R5

[Andreas Tobler]

Tom Rini wrote:
> 
> 
> What bug are you running into?  It's better to get 'em fixed now then try
> to work around 'em...
> 

When you have a look at page from John Lindal: http://www.cco.caltech.edu/~jafl/jx/jx2.html
There it explains the bug I ran/run into when I built JX.
Unfortunately there is no solution in sight from egcs people (as I understand)

Thanks

Andreas
-- 
------------------
| Andreas Tobler								
| CH-8004 Zuerich	
| E-mail: toa@pop.agri.ch		
------------------------------------------

[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to  Cc linuxppc-dev  if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]

Re: back to gcc under pre-R5

[Tom Rini]

On Wed, 17 Feb 1999, Andreas Tobler wrote:

> Tom Rini wrote:
> > 
> > 
> > What bug are you running into?  It's better to get 'em fixed now then try
> > to work around 'em...
> > 
> 
> When you have a look at page from John Lindal: http://www.cco.caltech.edu/~jafl/jx/jx2.html
> There it explains the bug I ran/run into when I built JX.
> Unfortunately there is no solution in sight from egcs people (as I understand)

Ah.  Well, either find out from John Lindal which files are effected by
this bug and compile those with -fno-vtable-thunks, or compile the whole
program with that.  Don't need to recompile libstdc++ or anything else.
(just checked with his testcase).

---
Tom Rini (TR1265)
http://dobbstown.yeti.edu/


[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to  Cc linuxppc-dev  if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]

Re: bugs building stuff.. somewhere.. (was: back to gcc under pre-R5)

[Brian Downing]

As a belated update to my own post:

I've finally gotten this to work.

"gcc -static -Wl,--relax ... -o binary" makes this link fine with
egcs-1.1.2 and glibc-2.1.  The resulting binary runs on pre-R5, R4,
and mklinux DR3.  I just wish I had a non-Mac to test it on.

-bcd
--
**               Brian Downing
**  UNIX Systems Administrator
**        bdowning@wolfram.com

On Wed, 17 Feb 1999 10:03:54 -0600, Brian Downing writes:
 > I've had similar problems building software.  I'm not sure where the
 > problem is, but I cannot compile this software statically.  I wouldn't
 > care, but for it to be useful, it must be able to run on all PPC
 > platforms.
 > 
 > When I link dynamically, everything works fine, everywhere that I've
 > built it - unfortunately binary portability is a little low.
 > 
 > When I try to link it statically, I get a bunch of ld errors from
 > libc.a about "tried to relocate __pthread_mutex_lock" and other various
 > __pthread_* calls.  No binary is generated.  I tried to compile a
 > non-reentrant version of glibc with no linuxthreads (since I'm building
 > statically and the software isn't threaded I figured it was worth a try).
 > Unfortunately glibc doesn't appear to build without linuxthreads anymore
 > - even with --disable-sanity-checks.  I've even gone so far as to ifdef
 > out the offending calls in glibc (for testing), and other similar errors
 > show up for different places in glibc.  I don't have the exact errors
 > on hand right now, but I can get them if anyone thinks they would help.

[snip]

 > (Built on a PowerBook G3 266, with linuxppc-pre-r5, with egcs-1.1.1 and
 > the latest snapshot I could build, with glibc-2.0.108 and glibc-2.1,
 > and with some other variations made for the heck of it.)

[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]