gcc for Irix.

gcc for Irix.

[Miguel de Icaza]

Hello guys,

   I am running into a little problem.  The binary gcc that is
available on the free software collection is for Irix 5.3 and the
include files that are packaged with it are not quite ok for Irix 6.2

   I do not have a native cc, so this is kind of problematic :-)

Miguel.

Re: gcc for Irix.

[William J. Earl]

Miguel de Icaza writes:
...
 >    I am running into a little problem.  The binary gcc that is
 > available on the free software collection is for Irix 5.3 and the
 > include files that are packaged with it are not quite ok for Irix 6.2
...

      In what way do the 5.3 include files cause problems?

Re: gcc for Irix.

[Ariel Faigon]

:
:Miguel de Icaza writes:
:...
: >    I am running into a little problem.  The binary gcc that is
: > available on the free software collection is for Irix 5.3 and the
: > include files that are packaged with it are not quite ok for Irix 6.2
:...
:
:      In what way do the 5.3 include files cause problems?
:

[Note: I'm taking care of Miguel in private email]

The headers that come with gcc are:

	1) Only a very small subset of the full set of headers
	   which were preprocessed by gcc 'fix_header' utility.

	2) Correspond to the OS version on which gcc was built
	   which may be incompatible in subtle ways with the
	   runtime libraries on a later OS version.

Since Miguel's Indy is (apparently) a 6.2 IRIX, possibly
without all the necessary header patches, and the gcc he
is trying was built on 5.3, I can see why he has header
problems.

Anyway, don't worry. This is being taken care of.
-- 
Peace, Ariel

Re: gcc for Irix.

[Ariel Faigon]

:
:Miguel de Icaza writes:
:...
: >    I am running into a little problem.  The binary gcc that is
: > available on the free software collection is for Irix 5.3 and the
: > include files that are packaged with it are not quite ok for Irix 6.2
:...
:
:      In what way do the 5.3 include files cause problems?
:

[Note: I'm taking care of Miguel in private email]

The headers that come with gcc are:

	1) Only a very small subset of the full set of headers
	   which were preprocessed by gcc 'fix_header' utility.

	2) Correspond to the OS version on which gcc was built
	   which may be incompatible in subtle ways with the
	   runtime libraries on a later OS version.

Since Miguel's Indy is (apparently) a 6.2 IRIX, possibly
without all the necessary header patches, and the gcc he
is trying was built on 5.3, I can see why he has header
problems.

Anyway, don't worry. This is being taken care of.
-- 
Peace, Ariel

Re: gcc for Irix.

[William J. Earl]

Ariel Faigon writes:
 > :
 > :Miguel de Icaza writes:
 > :...
 > : >    I am running into a little problem.  The binary gcc that is
 > : > available on the free software collection is for Irix 5.3 and the
 > : > include files that are packaged with it are not quite ok for Irix 6.2
 > :...
 > :
 > :      In what way do the 5.3 include files cause problems?
 > :
 > 
 > [Note: I'm taking care of Miguel in private email]
 > 
 > The headers that come with gcc are:
 > 
 > 	1) Only a very small subset of the full set of headers
 > 	   which were preprocessed by gcc 'fix_header' utility.
 > 
 > 	2) Correspond to the OS version on which gcc was built
 > 	   which may be incompatible in subtle ways with the
 > 	   runtime libraries on a later OS version.
 > 
 > Since Miguel's Indy is (apparently) a 6.2 IRIX, possibly
 > without all the necessary header patches, and the gcc he
 > is trying was built on 5.3, I can see why he has header
 > problems.

       If the old headers do not work on the new system, that is a
bug, not a feature.  There had better not be any subtle
incompatibilities.  (New headers will not work on an old system, but
that is a different problem.)

Re: gcc for Irix.

[William J. Earl]

Ariel Faigon writes:
 > :
 > :Miguel de Icaza writes:
 > :...
 > : >    I am running into a little problem.  The binary gcc that is
 > : > available on the free software collection is for Irix 5.3 and the
 > : > include files that are packaged with it are not quite ok for Irix 6.2
 > :...
 > :
 > :      In what way do the 5.3 include files cause problems?
 > :
 > 
 > [Note: I'm taking care of Miguel in private email]
 > 
 > The headers that come with gcc are:
 > 
 > 	1) Only a very small subset of the full set of headers
 > 	   which were preprocessed by gcc 'fix_header' utility.
 > 
 > 	2) Correspond to the OS version on which gcc was built
 > 	   which may be incompatible in subtle ways with the
 > 	   runtime libraries on a later OS version.
 > 
 > Since Miguel's Indy is (apparently) a 6.2 IRIX, possibly
 > without all the necessary header patches, and the gcc he
 > is trying was built on 5.3, I can see why he has header
 > problems.

       If the old headers do not work on the new system, that is a
bug, not a feature.  There had better not be any subtle
incompatibilities.  (New headers will not work on an old system, but
that is a different problem.)

Re: gcc for Irix.

[Christopher W. Carlson]

On Jun 13,  1:37pm, William J. Earl wrote:
> 
>        If the old headers do not work on the new system, that is a
> bug, not a feature.  There had better not be any subtle
> incompatibilities.  (New headers will not work on an old system, but
> that is a different problem.)
>-- End of excerpt from William J. Earl


Actually, it isn't a bug because there was a major version change.
Anything can change between 5.3 and 6.2.  I know we try to maintain
backwards compatibility but sometimes ANSI requires changes in the
number of parameters, there's a change in the values used to pass to
functions, etc.

I believe our promise is that code compiled on 5.3 will run on 6.2,
not that it will still compile.

-- 

		Chris Carlson

	+------------------------------------------------------+
	| Also, carlson@sgi.com                                |
	|   Work:   (714) 224-4530                             |
	|   Vnet:       6-678-4530     FAX:    (714) 833-9503  |
	|                                                      |
	| Trivia fact: an electroencephalogram shows that a    |
	| human brain and a bowl of quivering lime Jell-O have |
	| the same waves.  [Time Magazine, Mar 17, 1997]       |
	+------------------------------------------------------+

Re: gcc for Irix.

[Ariel Faigon]

:
:Actually, it isn't a bug because there was a major version change.
:Anything can change between 5.3 and 6.2.  I know we try to maintain
:backwards compatibility but sometimes ANSI requires changes in the
:number of parameters, there's a change in the values used to pass to
:functions, etc.
:
:I believe our promise is that code compiled on 5.3 will run on 6.2,
:not that it will still compile.
:
Yes, It was mainly ANSI/POSIX1C/POSIX2/XPG4/UNIX95 (spec 1170) changes
(See /usr/include/standards.h) which caused major changes in IRIX
headers in order to be able to support all these standards without
conflicts (a pretty big undertaking).

In addition there were some enhancemnets to certain APIs, for example
supporting faster serial speeds which changed some termios.h subtle
interfaces (you may still compile with -D_OLD_TERMIOS of course, but
still a change that may break compilations on old and arguably not
cleanly written UNIX apps -- GNU less comes to mind) 

And finally there's "fix_headers" - the utility that comes with gcc
and fixes headers so they can be used with some gcc conventions
and extensions to C.  Combine this with our multi-standard headers
which I suspect the designers of "fix_headers" never thought of
and you get a pretty cool mess :-)

-- 
Peace, Ariel

Re: gcc for Irix.

[Ariel Faigon]

:
:Actually, it isn't a bug because there was a major version change.
:Anything can change between 5.3 and 6.2.  I know we try to maintain
:backwards compatibility but sometimes ANSI requires changes in the
:number of parameters, there's a change in the values used to pass to
:functions, etc.
:
:I believe our promise is that code compiled on 5.3 will run on 6.2,
:not that it will still compile.
:
Yes, It was mainly ANSI/POSIX1C/POSIX2/XPG4/UNIX95 (spec 1170) changes
(See /usr/include/standards.h) which caused major changes in IRIX
headers in order to be able to support all these standards without
conflicts (a pretty big undertaking).

In addition there were some enhancemnets to certain APIs, for example
supporting faster serial speeds which changed some termios.h subtle
interfaces (you may still compile with -D_OLD_TERMIOS of course, but
still a change that may break compilations on old and arguably not
cleanly written UNIX apps -- GNU less comes to mind) 

And finally there's "fix_headers" - the utility that comes with gcc
and fixes headers so they can be used with some gcc conventions
and extensions to C.  Combine this with our multi-standard headers
which I suspect the designers of "fix_headers" never thought of
and you get a pretty cool mess :-)

-- 
Peace, Ariel

Re: gcc for Irix.

[David S. Miller]

   From: ariel@yon.engr.sgi.com (Ariel Faigon)
   Date: Sat, 14 Jun 1997 10:21:07 -0700 (PDT)

   And finally there's "fix_headers" - the utility that comes with gcc
   and fixes headers so they can be used with some gcc conventions and
   extensions to C.  Combine this with our multi-standard headers
   which I suspect the designers of "fix_headers" never thought of and
   you get a pretty cool mess :-)

Which is why in gcc-2.8.0 a completely new fix_headers.irix will exist
which handles it all properly...

Re: gcc for Irix.

[Martin Knoblauch]

David S. Miller wrote:
> 
>    From: ariel@yon.engr.sgi.com (Ariel Faigon)
>    Date: Sat, 14 Jun 1997 10:21:07 -0700 (PDT)
> 
>    And finally there's "fix_headers" - the utility that comes with gcc
>    and fixes headers so they can be used with some gcc conventions and
>    extensions to C.  Combine this with our multi-standard headers
>    which I suspect the designers of "fix_headers" never thought of and
>    you get a pretty cool mess :-)
> 
> Which is why in gcc-2.8.0 a completely new fix_headers.irix will
> exist which handles it all properly...

 This is why we all eagerly wait to see an official release of 2.8 :-)

Martin
-- 
+---------------------------------+-----------------------------------+
|Martin Knoblauch                 | Silicon Graphics GmbH             |
|Manager Technical Marketing      | Am Hochacker 3 - Technopark       |
|Silicon Graphics Computer Systems| D-85630 Grasbrunn-Neukeferloh, FRG|
|---------------------------------| Phone: (+int) 89 46108-179 or -0  |
|http://reality.sgi.com/knobi     | Fax:   (+int) 89 46107-179        |
+---------------------------------+-----------------------------------+
|e-mail: <knobi@munich.sgi.com>   | VM: 6-333-8197 | M/S: IDE-3150    |
+---------------------------------------------------------------------+

Re: gcc for Irix.

[Ralf Baechle]

> >    And finally there's "fix_headers" - the utility that comes with gcc
> >    and fixes headers so they can be used with some gcc conventions and
> >    extensions to C.  Combine this with our multi-standard headers
> >    which I suspect the designers of "fix_headers" never thought of and
> >    you get a pretty cool mess :-)
> > 
> > Which is why in gcc-2.8.0 a completely new fix_headers.irix will
> > exist which handles it all properly...
> 
>  This is why we all eagerly wait to see an official release of 2.8 :-)

Let things settle down a bit.  I tried to install a snapshot some days
ago and had problems to bootstrap the compiler.  The thin is still
pretty buggy.

  Ralf

Re: gcc for Irix.

[Ralf Baechle]

> >    And finally there's "fix_headers" - the utility that comes with gcc
> >    and fixes headers so they can be used with some gcc conventions and
> >    extensions to C.  Combine this with our multi-standard headers
> >    which I suspect the designers of "fix_headers" never thought of and
> >    you get a pretty cool mess :-)
> > 
> > Which is why in gcc-2.8.0 a completely new fix_headers.irix will
> > exist which handles it all properly...
> 
>  This is why we all eagerly wait to see an official release of 2.8 :-)

Let things settle down a bit.  I tried to install a snapshot some days
ago and had problems to bootstrap the compiler.  The thin is still
pretty buggy.

  Ralf

Re: IRIX ELF docs

[Christopher W. Carlson]

On Sep 30,  4:42pm, William J. Earl wrote:
> Subject: Re: IRIX ELF docs
> Ralf Baechle writes:
>  > Hi all,
>  >
>  > our current linker is producing IRIX flavored ELF binaries, not MIPS
>  > ABI.  We still seem to have some bugs in the dynamic linker and these
>  > are now pretty close to the top on my to do list.  However I've got
>  > not documentation about the IRIX binary format, so I'm pretty much
>  > relying on reverse engineering for fixing them.  Does anybody have
>  > a pointer to documentation or documentation about IRIX ELF flavoured
>  > o32 bit object file format?
>
>       IRIX ELF O32 (dynamic) object files are MIPS ABI object files.
> There are optional extra sections, to support features such as "quickstart"
> (which allows RLD to skip some of the fixups at startup time), but the
> required parts are as defined by the MIPS ABI.
>-- End of excerpt from William J. Earl


I recently went to the local technical bookstore and found that AT&T
no longer publishes the ABI books (the blue covered books describing
the ELF format).  The bookstore was, thus, unable to even order them
for me.

Does anybody know where to buy these books?

-- 

		Chris Carlson

	+------------------------------------------------------+
	| Also, carlson@sgi.com                                |
	|   Work:   (714) 224-4530                             |
	|   Vnet:       6-678-4530     FAX:    (714) 833-9503  |
	|                                                      |
	| Trivia fact: an electroencephalogram shows that a    |
	| human brain and a bowl of quivering lime Jell-O have |
	| the same waves.  [Time Magazine, Mar 17, 1997]       |
	+------------------------------------------------------+

Re: IRIX ELF docs

[Ariel Faigon]

Chris wrote:
:
:I recently went to the local technical bookstore and found that AT&T
:no longer publishes the ABI books (the blue covered books describing
:the ELF format).  The bookstore was, thus, unable to even order them
:for me.
:
:Does anybody know where to buy these books?
:

Not specifically on the ABI, but a search for "MIPS" on www.amazon.com
gives only 4 hits, two of which are out-of-print/hard-to-find and
probably irrelevant, and two that on the architecture only:

	Mips R4000 User's Manual
	Joseph Heinrich / Paperback / Published 1993 
	Our Price: $44.00 

	Mips Risc Architecture ~ Ships in 2-3 days
	Gerry Kane, Joe Heinrich / Paperback / Published 1991 
	Our Price: $46.00 
	Read more about this title... 

	Mips-X Risc Microprocessor (Kluwer International Series in
	Engineering and Computer Science) 
	Paul Chow / Hardcover / Published 1989 
	(Publisher Out Of Stock)

	Mips Risc Architecture 
	Published 1991 
	(Hard to Find)

The canonical place to look for MIPS ABI stuff is:
	www.mipsabi.org

The SGI specific quickstart stuff and whatever is not published
on www.mipsabi.org is probably not publicly documented and
I suspect not even documented much beyond the (rld etc.) source itself.
I'll be glad to be proven wrong on this.
-- 
Peace, Ariel

Re: IRIX ELF docs

[Ariel Faigon]

Chris wrote:
:
:I recently went to the local technical bookstore and found that AT&T
:no longer publishes the ABI books (the blue covered books describing
:the ELF format).  The bookstore was, thus, unable to even order them
:for me.
:
:Does anybody know where to buy these books?
:

Not specifically on the ABI, but a search for "MIPS" on www.amazon.com
gives only 4 hits, two of which are out-of-print/hard-to-find and
probably irrelevant, and two that on the architecture only:

	Mips R4000 User's Manual
	Joseph Heinrich / Paperback / Published 1993 
	Our Price: $44.00 

	Mips Risc Architecture ~ Ships in 2-3 days
	Gerry Kane, Joe Heinrich / Paperback / Published 1991 
	Our Price: $46.00 
	Read more about this title... 

	Mips-X Risc Microprocessor (Kluwer International Series in
	Engineering and Computer Science) 
	Paul Chow / Hardcover / Published 1989 
	(Publisher Out Of Stock)

	Mips Risc Architecture 
	Published 1991 
	(Hard to Find)

The canonical place to look for MIPS ABI stuff is:
	www.mipsabi.org

The SGI specific quickstart stuff and whatever is not published
on www.mipsabi.org is probably not publicly documented and
I suspect not even documented much beyond the (rld etc.) source itself.
I'll be glad to be proven wrong on this.
-- 
Peace, Ariel

Re: IRIX ELF docs

[Ralf Baechle]

> The SGI specific quickstart stuff and whatever is not published
> on www.mipsabi.org is probably not publicly documented and
> I suspect not even documented much beyond the (rld etc.) source itself.
> I'll be glad to be proven wrong on this.

Apropos quickstart, I recently had a discussion with someone about
quickstart & performance.  Does anybody have hard numbers about how
much the speedup by quickstart is?

  Ralf

Re: IRIX ELF docs

[Martin Knoblauch]

Ralf Baechle wrote:
> 
> 
> Apropos quickstart, I recently had a discussion with someone about
> quickstart & performance.  Does anybody have hard numbers about how
> much the speedup by quickstart is?
> 
>   Ralf
Ralf,

 one of these typical "it depends" questions :-) If your application
has only a short list of DSOs to load and if it runs for hours,
the effect of quickstart is pretty low. If your app loads a
lot of DSOs, which may in turn load other DSOs, the effect on
the startup time maybe pretty big (50%). And if the ratio between
startup-time and real-work-time is in favour of startup, you
better put some effort in quickstarting your DSOs.

 The major problem is, that the more DSOs you have, the more
difficult it is to get all of the quickstarting. So a lot of
people never do the work.

Martin
-- 
+---------------------------------+-----------------------------------+
|Martin Knoblauch                 | Silicon Graphics GmbH             |
|Manager Technical Marketing      | Am Hochacker 3 - Technopark       |
|Silicon Graphics Computer Systems| D-85630 Grasbrunn-Neukeferloh, FRG|
|---------------------------------| Phone: (+int) 89 46108-179 or -0  |
|http://reality.sgi.com/knobi     | Fax:   (+int) 89 46107-179        |
+---------------------------------+-----------------------------------+
|e-mail: <knobi@munich.sgi.com>   | VM: 6-333-8179 | M/S: IDE-3150    |
+---------------------------------------------------------------------+

Re: gcc for Irix.

[Martin Knoblauch]

Miguel de Icaza wrote:
> 
> Hello guys,
> 
>    I am running into a little problem.  The binary gcc that is
> available on the free software collection is for Irix 5.3 and the
> include files that are packaged with it are not quite ok for Irix 6.2
> 
>    I do not have a native cc, so this is kind of problematic :-)
> 
> Miguel.
Miguel,

 if you took the stuff from Ariels home page, don't worry
about the "5.3" stuff. It is basically the only way to build
an "official" gcc for 6.x. Otherwise, the stuff is binary
compatible. As for the header files, what is the problem?
Errors, or incompleteness?

Martin
-- 
+---------------------------------+-----------------------------------+
|Martin Knoblauch                 | Silicon Graphics GmbH             |
|Manager Technical Marketing      | Am Hochacker 3 - Technopark       |
|Silicon Graphics Computer Systems| D-85630 Grasbrunn-Neukeferloh, FRG|
|---------------------------------| Phone: (+int) 89 46108-179 or -0  |
|http://reality.sgi.com/knobi     | Fax:   (+int) 89 46107-179        |
+---------------------------------+-----------------------------------+
|e-mail: <knobi@munich.sgi.com>   | VM: 6-333-8197 | M/S: IDE-3150    |
+---------------------------------------------------------------------+

IRIX ELF docs

[Ralf Baechle]

Hi all,

our current linker is producing IRIX flavored ELF binaries, not MIPS
ABI.  We still seem to have some bugs in the dynamic linker and these
are now pretty close to the top on my to do list.  However I've got
not documentation about the IRIX binary format, so I'm pretty much
relying on reverse engineering for fixing them.  Does anybody have
a pointer to documentation or documentation about IRIX ELF flavoured
o32 bit object file format?

  Ralf

Re: IRIX ELF docs

[William J. Earl]

Ralf Baechle writes:
 > Hi all,
 > 
 > our current linker is producing IRIX flavored ELF binaries, not MIPS
 > ABI.  We still seem to have some bugs in the dynamic linker and these
 > are now pretty close to the top on my to do list.  However I've got
 > not documentation about the IRIX binary format, so I'm pretty much
 > relying on reverse engineering for fixing them.  Does anybody have
 > a pointer to documentation or documentation about IRIX ELF flavoured
 > o32 bit object file format?

      IRIX ELF O32 (dynamic) object files are MIPS ABI object files.
There are optional extra sections, to support features such as "quickstart"
(which allows RLD to skip some of the fixups at startup time), but the
required parts are as defined by the MIPS ABI.