[parisc-linux] crti.S Nested procedures error

[parisc-linux] crti.S Nested procedures error

[Grant Grundler]

Sammy,

I tried the "glibc" recipe on my x86 box (running Debian potato release).
(http://puffin.external.hp.com/mailing-lists/parisc-linux/2000/04-Apr/0059.html)

"make" died with the following error:

...
hppa1.1-linux-gcc /linux/grundler/glibc/hppa1.1-linux/csu/crti.S -c  -I../include -I. -I/linux/grundler/glibc/hppa1.1-linux/csu -I.. -I../libio  -I/linux/grundler/glibc/hppa1.1-linux -I../sysdeps/hppa/elf -I../sysdeps/unix/sysv/linux/hppa -I../sysdeps/unix/sysv/linux -I../sysdeps/gnu -I../sysdeps/unix/common -I../sysdeps/unix/mman -I../sysdeps/unix/inet -I../sysdeps/unix/sysv -I../sysdeps/unix -I../sysdeps/posix -I../sysdeps/hppa/hppa1.1 -I../sysdeps/wordsize-32 -I../sysdeps/ieee754/flt-32 -I../sysdeps/ieee754/dbl-64 -I../sysdeps/ieee754/ldbl-128 -I../sysdeps/hppa/fpu -I../sysdeps/hppa -I../sysdeps/ieee754 -I../sysdeps/generic/elf -I../sysdeps/generic  -nostdinc -isystem /opt/puffin/lib/gcc-lib/hppa1.1-linux/2.9-hppa-991112/include -isystem /opt/puffin/include  -include ../include/libc-symbols.h     -DHAVE_INITFINI -DASSEMBLER  -I/linux/grundler/glibc/hppa1.1-linux/csu/.  -g0  -o /linux/grundler/glibc/hppa1.1-linux/csu/crti.o

bc-symbols.h     -DHAVE_INITFINI -DASSEMBLER  -I/linux/grundler/glibc/hppa1.1-linux/csu/.  -g0  -o /linux/grundler/glibc/hppa1.1-linux/csu/crti.o
/linux/grundler/glibc/hppa1.1-linux/csu/crti.S: Assembler messages:
/linux/grundler/glibc/hppa1.1-linux/csu/crti.S:42: Fatal error: Nested procedures
make[2]: *** [/linux/grundler/glibc/hppa1.1-linux/csu/crti.o] Error 1
make[2]: Leaving directory `/linux/grundler/glibc/csu'
make[1]: *** [csu/subdir_lib] Error 2


Sure enough, two .PROC's have no .PROCEND in between.
glibc/csu/Makefile builds crti.S with:

$(objpfx)crti.S: $(objpfx)initfini.s
        sed -n -e '1,/@HEADER_ENDS/p' \
               -e '/@_.*_PROLOG_BEGINS/,/@_.*_PROLOG_ENDS/p' \
               -e '/@TRAILER_BEGINS/,$$p' $< > $@ 


It seems either the initfini.s contents are wrong or the "sed" hack
above is removing too much.  Which is it?
Or should there only be one .PROC?

We get two .PROC's because initfini.s has both @_init_PROLOG_BEGINS/ENDS
and @_fini_PROLOG_BEGINS/ENDS. 

Appended "diff -u initfini.s crti.S" shows what sed is doing.

thanks,
grant


grundler <576>diff -u initfini.s crti.S
--- initfini.s  Mon May  1 11:09:43 2000
+++ crti.S      Mon May  1 11:20:48 2000
@@ -6,38 +6,6 @@
 #include "defs.h"
 
 /*@HEADER_ENDS*/
-
-/*@TESTS_BEGIN*/
-#NO_APP
-       .text
-       .align 4
-.globl dummy
-       .EXPORT dummy,ENTRY
-       .type    dummy,@function
-dummy:
-dummy:
-       .PROC
-       .CALLINFO FRAME=64,CALLS,SAVE_RP,ENTRY_GR=3
-       .ENTRY
-       stw %r2,-20(%r30)
-       ldo 64(%r30),%r30
-       stw %r19,-32(%r30)
-       stw %r4,-64(%r30)
-       movb,= %r26,%r22,.L16
-       copy %r19,%r4
-       .CALL   ARGW0=GR
-       bl $$dyncall,%r31
-       copy %r31,%r2
-.L16:
-       ldw -84(%r30),%r2
-       bv %r0(%r2)
-       ldwm -64(%r30),%r4
-       .EXIT
-       .PROCEND
-#APP
-
-/*@TESTS_END*/
-
 /*@_init_PROLOG_BEGINS*/
        .section .init
 #NO_APP
@@ -62,35 +30,6 @@
        END_INIT
 
 /*@_init_PROLOG_ENDS*/
-
-/*@_init_EPILOG_BEGINS*/
-       .section .init
-#NO_APP
-       ldw -84(%r30),%r2
-       bv %r0(%r2)
-       ldwm -64(%r30),%r4
-       .EXIT
-       .PROCEND
-#APP
-       END_INIT
-       .section .text
-#NO_APP
-       .align 4
-       .weak   __gmon_start__
-       .EXPORT __gmon_start__,ENTRY
-       .type    __gmon_start__,@function
-__gmon_start__:
-__gmon_start__:
-       .PROC
-       .CALLINFO FRAME=0,NO_CALLS
-       .ENTRY
-       bv,n %r0(%r2)
-       .EXIT
-       .PROCEND
-#APP
-
-/*@_init_EPILOG_ENDS*/
-
 /*@_fini_PROLOG_BEGINS*/
        .section .fini
 #NO_APP
@@ -113,24 +52,5 @@
        END_FINI
 
 /*@_fini_PROLOG_ENDS*/
-#NO_APP
-       .CALL 
-       bl i_am_not_a_leaf,%r2
-       nop
-#APP
-
-/*@_fini_EPILOG_BEGINS*/
-       .section .fini
-#NO_APP
-       ldw -84(%r30),%r2
-       bv %r0(%r2)
-       ldwm -64(%r30),%r4
-       .EXIT
-       .PROCEND
-#APP
-       END_FINI
-
-/*@_fini_EPILOG_ENDS*/
-
 /*@TRAILER_BEGINS*/
        .ident  "GCC: (GNU) 2.9-hppa-991112"

Re: [parisc-linux] crti.S Nested procedures error

[Sam Creasey]

On Mon, 1 May 2000, Grant Grundler wrote:

> Sammy,
> 
> I tried the "glibc" recipe on my x86 box (running Debian potato release).
> (http://puffin.external.hp.com/mailing-lists/parisc-linux/2000/04-Apr/0059.html)
> 
> "make" died with the following error:
> 
> ...
> hppa1.1-linux-gcc /linux/grundler/glibc/hppa1.1-linux/csu/crti.S -c  -I../include -I. -I/linux/grundler/glibc/hppa1.1-linux/csu -I.. -I../libio  -I/linux/grundler/glibc/hppa1.1-linux -I../sysdeps/hppa/elf -I../sysdeps/unix/sysv/linux/hppa -I../sysdeps/unix/sysv/linux -I../sysdeps/gnu -I../sysdeps/unix/common -I../sysdeps/unix/mman -I../sysdeps/unix/inet -I../sysdeps/unix/sysv -I../sysdeps/unix -I../sysdeps/posix -I../sysdeps/hppa/hppa1.1 -I../sysdeps/wordsize-32 -I../sysdeps/ieee754/flt-32 -I../sysdeps/ieee754/dbl-64 -I../sysdeps/ieee754/ldbl-128 -I../sysdeps/hppa/fpu -I../sysdeps/hppa -I../sysdeps/ieee754 -I../sysdeps/generic/elf -I../sysdeps/generic  -nostdinc -isystem /opt/puffin/lib/gcc-lib/hppa1.1-linux/2.9-hppa-991112/include -isystem /opt/puffin/include  -include ../include/libc-symbols.h     -DHAVE_INITFINI -DASSEMBLER  -I/linux/grundler/glibc/hppa1.1-linux/csu/.  -g0  -o /linux/grundler/glibc/hppa1.1-linux/csu/crti.o
> 
> bc-symbols.h     -DHAVE_INITFINI -DASSEMBLER  -I/linux/grundler/glibc/hppa1.1-linux/csu/.  -g0  -o /linux/grundler/glibc/hppa1.1-linux/csu/crti.o
> /linux/grundler/glibc/hppa1.1-linux/csu/crti.S: Assembler messages:
> /linux/grundler/glibc/hppa1.1-linux/csu/crti.S:42: Fatal error: Nested procedures
> make[2]: *** [/linux/grundler/glibc/hppa1.1-linux/csu/crti.o] Error 1
> make[2]: Leaving directory `/linux/grundler/glibc/csu'
> make[1]: *** [csu/subdir_lib] Error 2
> 
> 
> Sure enough, two .PROC's have no .PROCEND in between.
> glibc/csu/Makefile builds crti.S with:
> 
> $(objpfx)crti.S: $(objpfx)initfini.s
>         sed -n -e '1,/@HEADER_ENDS/p' \
>                -e '/@_.*_PROLOG_BEGINS/,/@_.*_PROLOG_ENDS/p' \
>                -e '/@TRAILER_BEGINS/,$$p' $< > $@ 
> 
> 
> It seems either the initfini.s contents are wrong or the "sed" hack
> above is removing too much.  Which is it?
> Or should there only be one .PROC?
> 
> We get two .PROC's because initfini.s has both @_init_PROLOG_BEGINS/ENDS
> and @_fini_PROLOG_BEGINS/ENDS. 
> 
> Appended "diff -u initfini.s crti.S" shows what sed is doing.
> 
> thanks,
> grant

ok, for a while I was dealing with this by simply hacking up initfini.s by
hand, but that wasn't necessary last time I built.  I think this is
related to the fact that somebody had taken out .proc/.procend from the
assembler entirely (well, ignored them  on ELF targets).  I think this was
a bad idea (if I remember tc-hppa.c from gas, it depends on these to build
some of the global symbol information?  It did something even on an elf
target).

We might need our own magic sed script for parisc.

Lemme think about this properly, I just woke up and am expected at work.

-- Sam

Re: [parisc-linux] crti.S Nested procedures error

[willy]

On Wed, May 03, 2000 at 10:18:54AM -0400, Sam Creasey wrote:
> ok, for a while I was dealing with this by simply hacking up initfini.s by
> hand, but that wasn't necessary last time I built.  I think this is
> related to the fact that somebody had taken out .proc/.procend from the
> assembler entirely (well, ignored them  on ELF targets).  I think this was
> a bad idea (if I remember tc-hppa.c from gas, it depends on these to build
> some of the global symbol information?  It did something even on an elf
> target).

it was a bad idea, it broke other things.  i suspect the right thing
to do is move some of the global symbol information-building code
which is currently invoked upon seeing a .proc or .procend to simply
seeing a non-local label, like the other assemblers do.  i don't see why
linux/elf/pa-risc needs anything special in this regard.  Alan (Modra),
do you have any comments?

> We might need our own magic sed script for parisc.

i'd prefer not to..

Re: [parisc-linux] crti.S Nested procedures error

[Alan Modra]

On Wed, 3 May 2000 willy@thepuffingroup.com wrote:

> On Wed, May 03, 2000 at 10:18:54AM -0400, Sam Creasey wrote:
> > ok, for a while I was dealing with this by simply hacking up initfini.s by
> > hand, but that wasn't necessary last time I built.  I think this is
> > related to the fact that somebody had taken out .proc/.procend from the
> > assembler entirely (well, ignored them  on ELF targets).  I think this was
> > a bad idea (if I remember tc-hppa.c from gas, it depends on these to build
> > some of the global symbol information?  It did something even on an elf
> > target).
> 
> it was a bad idea, it broke other things.  i suspect the right thing
> to do is move some of the global symbol information-building code
> which is currently invoked upon seeing a .proc or .procend to simply
> seeing a non-local label, like the other assemblers do.  i don't see why
> linux/elf/pa-risc needs anything special in this regard.  Alan (Modra),
> do you have any comments?

Well, I'm still coming up to speed on the parisc architecture, so maybe my
comments aren't that valuable.  There are quite a lot of things I don't
like about the current set of puffin parisc patches though, and zapping
these pseudo ops is one of them.

On another issue, can someone tell me why the assembler comment character
needed to be changed from `;' to `!' ?  This seems a really ill-advised
change to me.

-- 
Linuxcare.  Support for the Revolution.

Re: [parisc-linux] crti.S Nested procedures error

[Alan Modra]

On Wed, 3 May 2000, Grant Grundler wrote:

> Alan Modra wrote:
> 
> > On another issue, can someone tell me why the assembler comment character
> > needed to be changed from `;' to `!' ?  This seems a really ill-advised
> > change to me.
> 
> See:
> http://puffin.external.hp.com/mailing-lists/parisc-linux/2000/03-Mar/0078.html

Thanks for the pointer.  I agree with all these except the comment
change...  :)

-- 
Linuxcare.  Support for the Revolution.

Re: [parisc-linux] crti.S Nested procedures error

[Jeffrey A Law]

  In message <Pine.LNX.4.21.0005041047580.8334-100000@front.linuxcare.com.au>yo
u write:
  > On Wed, 3 May 2000, Grant Grundler wrote:
  > 
  > > Alan Modra wrote:
  > > 
  > > > On another issue, can someone tell me why the assembler comment charact
  > er
  > > > needed to be changed from `;' to `!' ?  This seems a really ill-advised
  > > > change to me.
  > > 
  > > See:
  > > http://puffin.external.hp.com/mailing-lists/parisc-linux/2000/03-Mar/0078
  > .html
  > 
  > Thanks for the pointer.  I agree with all these except the comment
  > change...  :)
'!' is the standard comment character for the PA assembly syntax.

jeff

Re: [parisc-linux] crti.S Nested procedures error

[Alan Modra]

On Wed, 3 May 2000, Jeffrey A Law wrote:

> '!' is the standard comment character for the PA assembly syntax.

sourceware.cygnus.com CVS gas/config/tc-hppa.c has

const char comment_chars[] = ";";
const char line_separator_chars[] = "!";

You're saying that it's a good idea to change the comment char then?

Regards, Alan Modra

-- 
Linuxcare.  Support for the Revolution.

Re: [parisc-linux] crti.S Nested procedures error

[Jeffrey A Law]

  In message <Pine.LNX.4.21.0005041123510.8334-100000@front.linuxcare.com.au>yo
u write:
  > On Wed, 3 May 2000, Jeffrey A Law wrote:
  > 
  > > '!' is the standard comment character for the PA assembly syntax.
  > 
  > sourceware.cygnus.com CVS gas/config/tc-hppa.c has
  > 
  > const char comment_chars[] = ";";
  > const char line_separator_chars[] = "!";
  > 
  > You're saying that it's a good idea to change the comment char then?
Whoops.  My bad.  "!" is the line separator, not the comment character.

Sometimes I wonder if I'm going senile.

jeff

ps.  One can certainly argue about whether or not "!" is a good line
separator since it's also used in expression parsing and as a completer
for FP comparisons.  Basically we've got an ambiguous grammar due to some
poor choices that were made long ago ;(

jeff

Re: [parisc-linux] crti.S Nested procedures error

[Paul Bame]

= "make" died with the following error:
= 
= /linux/grundler/glibc/hppa1.1-linux/csu/crti.S: Assembler messages:
= /linux/grundler/glibc/hppa1.1-linux/csu/crti.S:42: Fatal error: Nested proced
ures
= 
= Sure enough, two .PROC's have no .PROCEND in between.

I think I've figured this out, and it'll take a glibc expert to do
the *right* thing vis-a-vis the complex glibc structure.

Let's ignore the .fini section and just deal with .init.  This is the
section where code which needs to execute prior to main() is placed.  It
is in effect a normal assembly-language procedure whose contents are
created at link-time.  The procedure looks like this (somewhat simplified):

    .globl _init
	    .EXPORT _init,ENTRY
	    .type    _init,@function
    _init:
	    .PROC
	    .CALLINFO FRAME=64,CALLS,SAVE_RP,ENTRY_GR=3
	    .ENTRY
	    a few assembly statements

	    /***** pre-main() code to be inserted here *******/

	    a few more assembly statements
	    .EXIT
	    .PROCEND

glibc wants to place the first part (up to the comment) into crti.S, and
the second part into crtn.S, then at link time something like this happens:

	ld -o a.out crti.o users's-object-files crtn.o

Inside the user's object files, if they need something called before main(),
like a C++ static constructor, you might find code like this:

	.section .init
	bl,n my-static-constructor,%r2

Thus the _init procedure is created is link time, but that's not the problem.

The problem is that the neither the first scrap of code nor the second
will assemble, because there's a .PROC without .PROCEND and .ENTRY
without .EXIT.  .ENTRY and .EXIT both produce code, .ENTRY depends on
.CALLINFO which depends a tad upon .PROC.

If we hand-code the .ENTRY and .EXIT sequences, we can nuke all the
troublesome assembler directives and fix this... simple!...

except for the way this code got generated in the first place.
It's generated in glibc/csu by 'gcc -S initfini.c' and then
hacking the generated assembly, which contains the .PROC,.ENTRY,etc...

So one thing we could do is hack gcc to produce its own entry/exit
assembly code generation.  Another thing we could do is make the
assembler very forgiving.  Perhaps the best thing we could do is find a way
to insert PA-specific handwritten assembly for these two files
into the glibc build.  I can't figure out how to do this without
making glibc specific only to PA.  Help!

	-Paul Bame

Re: [parisc-linux] crti.S Nested procedures error

[Cary Coutant]

Have you tried this:

>    .globl _init
>	    .EXPORT _init,ENTRY
>	    .type    _init,@function
>    _init:
>	    .PROC
>	    .CALLINFO FRAME=64,CALLS,SAVE_RP,ENTRY_GR=3
>	    .ENTRY
>	    a few assembly statements

      .PROCEND

      .PROC
 	    .CALLINFO FRAME=64,CALLS,SAVE_RP,ENTRY_GR=3
      
>
>	    /***** pre-main() code to be inserted here *******/
>
>	    a few more assembly statements
>	    .EXIT
>	    .PROCEND

While you need to use .proc and .procend in pairs, that's not the case 
with .entry and .exit. If the assembler won't take this code, it should. 
Besides being useful for this case, it's also necessary to generate code 
like this for out-of-line pieces of procedures.

-cary