Re: 970825 Snapshot Available

Re: 970825 Snapshot Available

[H.J. Lu]

> We've actually only tested with the full snapshot; we're interested in
> hearing if people can build the core or the core + particular language
> add-ons.
> 

What happened to my SMP patch? The build cannot be done without it
on an SMP machine.

--- 
H.J. Lu (hjl@gnu.ai.mit.edu)

Re: 970825 Snapshot Available

[Jeffrey A Law]

  In message <m0x37mv-0004ecC@ocean.lucon.org>you write:
  > What happened to my SMP patch? The build cannot be done without it
  > on an SMP machine.
It's in the queue.

However, you sent a patch without describing in any way, shape or
form what problem you were trying to solve.

As a result, it takes much longer to evaluate whether the patch
is necessary and correct -- which certainly puts it behind more
pressing issues in my queue like code generation and making single
threaded make work.

Maybe if you resent the patch and described the problem you were
trying to solve and your solution it would be more likely to go
into the tree in a timely fashion.

jeff

Re: -g1 lossage

[Robert Lipe]

> I did write a patch for this when it was reported to gcc2.  However, I did
> not test it.  I was hoping someone would report back to me whether it did
> work, but I never heard anything.  If someone tells me that it does work,
> then I will install it.

This would be my fault.  Sorry. I reported it just before 
effectively leaving town for a month.

I just applied it to today's egcs and it now no longer chokes 
when building new2.cc.   I haven't yet run the regressions on it,
but at least it builds.

Thanx for your help, Jim.

RJL


> Tue Aug  5 22:45:11 1997  Jim Wilson  <wilson@cygnus.com>
> 
> 	* dwarfout.c (dwarfout_file_scope_decl, case TYPE_DECL): Check
> 	TYPE_DECL_IS_STUB instead of DECL_NAME.
> 
> *** clean-testgcc-970803/dwarfout.c	Sat Aug  2 11:32:44 1997
> --- testgcc-970803/dwarfout.c	Tue Aug  5 22:29:46 1997
> *************** dwarfout_file_scope_decl (decl, set_fina
> *** 5144,5150 ****
>   	 a return type or a formal parameter type of some function.  */
>   
>         if (debug_info_level <= DINFO_LEVEL_TERSE)
> ! 	if (DECL_NAME (decl) != NULL
>   	    || ! TYPE_USED_FOR_FUNCTION (TREE_TYPE (decl)))
>             return;
>   
> --- 5144,5150 ----
>   	 a return type or a formal parameter type of some function.  */
>   
>         if (debug_info_level <= DINFO_LEVEL_TERSE)
> ! 	if (! TYPE_DECL_IS_STUB (decl)
>   	    || ! TYPE_USED_FOR_FUNCTION (TREE_TYPE (decl)))
>             return;
>   
> 
> 
> 

-- 
Robert Lipe       http://www.dgii.com/people/robertl       robertl@dgii.com

this week's patch for libgcc2-elf.a

[Robert Lipe]

This patch was originally against gcc2-970803, but it seems to 
vastly help egcs-970825.   Without it, libgcc2-elf.a doesn't 
contain any of the g++ stuff such as rtti.

I think there may still be a dependency missing, though.

Sun Aug  3 23:37:20 1997  Robert Lipe  <robertl@dgii.com>

	* config/i386/t-sco5: Resync libgcc2-elf.a target instructions 
		with libgcc2.a.

*** config/i386/t-sco5_	Fri Feb 14 14:41:31 1997
--- config/i386/t-sco5	Mon Jul 28 10:28:07 1997
***************
*** 13,18 ****
--- 13,23 ----
  # NOTE: This is duplicated directly from Makefile.in. If you change it
  # there, be sure to change things here!
  #
+ # NOTE: 07/28/97 robertl - this doesn't happen.   Makefile.in gets changed
+ #       and this file gets ignored.    This whole thing should either be
+ #       ignored (build it all as either ELF or COFF and let the SCO linker
+ #       convert things) or converted to the new multilib stuff.
+ #
  EXTRA_PARTS = libgcc-elf.a crtbegin.o crtend.o crtbeginS.o crtendS.o
  TAROUTOPTS = xpf
  
***************
*** 47,55 ****
  	-if $(RANLIB_TEST) ; then $(RANLIB) tmplibgcc1-elf.a; else true; fi
  	mv tmplibgcc1-elf.a libgcc1-elf.a
  
! libgcc2-elf.a: stamp-crtS libgcc2.c libgcc2.ready $(CONFIG_H) $(LIB2FUNCS_EXTRA) \
!    machmode.h longlong.h gbl-ctors.h config.status
  	-rm -f tmplibgcc2-elf.a
  	for name in $(LIB2FUNCS); \
  	do \
  	  echo $${name}; \
--- 52,68 ----
  	-if $(RANLIB_TEST) ; then $(RANLIB) tmplibgcc1-elf.a; else true; fi
  	mv tmplibgcc1-elf.a libgcc1-elf.a
  
! libgcc2-elf.a: libgcc2.c libgcc2.ready $(CONFIG_H) $(LIB2FUNCS_EXTRA) \
!    $(LANG_LIB2FUNCS) machmode.h longlong.h gbl-ctors.h config.status
! # Actually build it in tmplibgcc2.a, then rename at end,
! # so that libgcc2.a itself remains nonexistent if compilation is aborted.
  	-rm -f tmplibgcc2-elf.a
+ # -e causes any failing command to make this rule fail.
+ # -e doesn't work in certain shells, so we test $$? as well.
+ # lynx has a broken ar, it always complains when the initial library is
+ # empty, thus this command works only if we don't do -e
+ # There is a trailing backslash (\) deleted from the following line.
+ #	set -e;
  	for name in $(LIB2FUNCS); \
  	do \
  	  echo $${name}; \
***************
*** 59,81 ****
  	  $(AR) $(AR_FLAGS) tmplibgcc2-elf.a $${name}$(objext); \
  	  rm -f $${name}$(objext); \
  	done
! 	for file in .. $(LIB2FUNCS_EXTRA); \
  	do \
  	  if [ x$${file} != x.. ]; then \
! 	    name=`echo $${file} | sed -e 's/[.][cS]$$//' -e 's/[.]asm$$//'`; \
  	    oname=` echo $${name} | sed -e 's,.*/,,'`; \
! 	    echo $${name}; \
! 	    if [ $${name}.asm = $${file} ]; then \
! 	      cp $${file} $${name}.s || exit 1; file=$${name}.s; \
! 	    else true; fi; \
! 	    $(GCC_FOR_TARGET) -melf $(LIBGCC2_CFLAGS) $(INCLUDES) -c $${file}; \
! 	    if [ $$? -eq 0 ] ; then true; else exit 1; fi; \
! 	    $(AR) $(AR_FLAGS) tmplibgcc2-elf.a $${oname}$(objext); \
! 	    rm -f $${name}.s $${oname}$(objext); \
  	  else true; \
  	  fi; \
  	done
  	mv tmplibgcc2-elf.a libgcc2-elf.a
  
  # Combine the various libraries into a single library, libgcc.a.
  libgcc-elf.a: libgcc1-elf.a libgcc2-elf.a
--- 72,118 ----
  	  $(AR) $(AR_FLAGS) tmplibgcc2-elf.a $${name}$(objext); \
  	  rm -f $${name}$(objext); \
  	done
! # Some shells crash when a loop has no items.
! # So make sure there is always at least one--`..'.
! # Then ignore it.
! # We don't use -e here because there are if statements
! # that should not make the command give up when the if condition is false.
! # Instead, we test for failure after each command where it matters.
! 	for file in .. $(LIB2FUNCS_EXTRA) $(LANG_LIB2FUNCS); \
  	do \
  	  if [ x$${file} != x.. ]; then \
! 	    name=`echo $${file} | sed -e 's/[.][cSo]$$//' -e 's/[.]asm$$//' -e 's/[.]txt$$//'`; \
  	    oname=` echo $${name} | sed -e 's,.*/,,'`; \
! 	    if [ $${name}.txt = $${file} ]; then \
! 	      for f in .. `cat $${file}`; do if [ x$${f} != x.. ]; then \
! 		$(MAKE) GCC_FOR_TARGET="$(GCC_FOR_TARGET)" \
! 	  	  AR="$(AR)" AR_FLAGS="$(AR_FLAGS)" CC="$(CC)" \
! 		  CFLAGS="$(CFLAGS) " HOST_PREFIX="$(HOST_PREFIX)" \
! 		  HOST_PREFIX_1="$(HOST_PREFIX_1)" \
! 		  LIBGCC2_CFLAGS="$(LIBGCC2_CFLAGS) -melf" $${f}; \
! 	        if [ $$? -eq 0 ] ; then true; else exit 1; fi; \
! 	        $(AR) $(AR_FLAGS) tmplibgcc2-elf.a $${f}; \
! 	        rm -f $${f}; \
! 	      fi; done; \
! 	    else \
! 	      echo $${name}; \
! 	      if [ $${name}.asm = $${file} ]; then \
! 		cp $${file} $${name}.s || exit 1; file=$${name}.s; \
! 	      else true; fi; \
! 	      $(GCC_FOR_TARGET) -melf $(LIBGCC2_CFLAGS) $(INCLUDES) -c $${file}; \
! 	      if [ $$? -eq 0 ] ; then true; else exit 1; fi; \
! 	      $(AR) $(AR_FLAGS) tmplibgcc2-elf.a $${oname}$(objext); \
! 	      rm -f $${name}.s $${oname}$(objext); \
! 	    fi; \
  	  else true; \
  	  fi; \
  	done
  	mv tmplibgcc2-elf.a libgcc2-elf.a
+ # These lines were deleted from above the mv command
+ # because ranlibing libgcc.a itself should suffice.
+ #	-if [ x${HPUX_GAS} = x ] ; then \
+ #	  if $(RANLIB_TEST) ; then $(RANLIB) tmplibgcc2.a; else true; fi; \
+ #	else true; fi
  
  # Combine the various libraries into a single library, libgcc.a.
  libgcc-elf.a: libgcc1-elf.a libgcc2-elf.a

Suggestion re: 970825 Snapshot Available

[Julian Seward]

Jeffrey A Law wrote:
> 
> egcs-ss-970825 is now available in ftp.cygnus.com:/pub/egcs/snapshots
> (and on various mirrors shortly, see the home page for mirror sites).
[...]

I built egcs-ss-970821 for i486-linux-elf for c only, and it seems to
work, as far as my superficial tests can determine.  It's good to see
further development of gcc2.

I wondered if I could interest you in the following.  I imagine a lot
of people will be interested in trying egcs.  People with slow (modem
only) connections, like me, have some difficulty downloading egcs.  I
only have a 14.4k modem, and it took nearly two hours.  I guess many
people will have 28.8k or better modems, but it still takes on the
order of an hour.

Would you consider compressing egcs with a better compressor?  The
question would be a bit academic but for the fact that I can offer you
such a beast.  It's GNU GPL'd, patent-free, reasonably fast, portable,
and command-line compatible with gzip.

It saves nearly two megabytes for egcs:

   -rw-r--r--   37304320 Aug 25 12:04 egcs-ss-970821.tar
   -rw-r--r--    8043075 Aug 25 15:57 egcs-ss-970821.tar.gz
   -rw-r--r--    6169539 Aug 25 12:04 egcs-ss-970821.tar.bz2

For me that's around half an hour less on the phone, and also less
bandwidth on your servers.  Of course, people would have to download a
copy of the compressor, but that's about 300k for the source distrib,
or ~100k for a binary, so it would be a win even on the first
download.

The compressor's called "bzip2"; an early version of it has seen
some use in the Linux community over the past year.  You can
get it from http://www.muraroa.demon.co.uk .

Probably a rather strange-sounding proposition, but a serious one.

Julian Seward

Much c-torture success on 970825 :-)

[J. Kean Johnston]

Hi again.

On SCO Open Server 5.0, in both COFF and ELF modes, egcs 970825 passes
c-torture with flying colours. It only has a single expected failure.
I am about to do all of the libg++ tests, and will keep you all informed
of success on that front.

Also of interest, it is currently building cleanly on SCO UnixWare SVR5.
This is a HUGE win for the EGCS project. Well done everyone, this is
great. Just for the sake of satisfying my perverse curiosity, I am
going to try and build XFree86 3.3.1 with the newly built, fresh out
of the box egcs, and see what happens.

JKJ

Re: 970825 Snapshot Available

[Jim Wilson]

	What happened to my SMP patch?

There have so many people reporting bugs, submitting patches, etc, that it has
been impossible to handle them all in a timely fashion.  I expect that things
will get better soon, once we get past the initial surge caused by the project
startup.  There is also the problem that I am new at this, and still trying
to figure out how to deal with all of the email, prioritize work, etc.

So far, I have been concentrating on the build problems.  I wanted to get
as many as possible of the serious build problems fixed early, so that more
people would be able to try egcs.  This is also important since we are
planning on making an initial release sometime within the next few weeks.

I just today started looking at the various patches that have been submitted
since the project started.

I am already working on your SMP patch.  I expect to check it in today
unchanged.  My only comment is to notice that the f/runtime Makefile is
missing dependencies.  If one of the lib*77 libraries is updated, libf2c.a
won't be updated.  There were some partial dependencies for this before your
change, but they were flawed, so I does appear that they should be removed.
Still, it seems that this should be fixed someday.

	The build cannot be done without it on an SMP machine.

I wasn't aware of that.  I thought it was a problem only if you used the
-j option.  If -jN N!=1 is the default, can't you simply work around this
by using -j1?

Jim

Re: fixing the c++/f77 circular dependency

[Jim Wilson]

	Could someone please tell me why we cannot put a target libgcc.a in
	cp/Makefile.in? It will resolve that problem. I am willing to
	write a patch if people think it is a good idea.

There is a serious dependency problem which has already been mentioned a few
times.  If libgcc.a contains C++ code, then libgcc.a depends on cc1plus,
which means we always build cc1plus even if it is not in languages.  Simply
moving the rule to cp/Makefile.in does not solve this problem, because
we still have libgcc.a dependencing on cc1plus.

The existing rule for the C++ libgcc.a stuff uses contortions to try to
avoid an explicit cc1plus dependency, but it is ugly, does not work reliably,
and has been causing problems.

Jim

970825 Snapshot Available

[Jeffrey A Law]

egcs-ss-970825 is now available in ftp.cygnus.com:/pub/egcs/snapshots
(and on various mirrors shortly, see the home page for mirror sites).

You'll find:

  egcs-970825.tar.gz                    The full egcs snapshot, including all
                                        languages runtime libraries and
                                        testsuites.

  egcs-core-970825.tar.gz               Just the C and objc language front ends.

  egcs-g++-970825.tar.gz                The g++ language and runtime.

  egcs-g77-970825.tar.gz                The g77 language and runtime.
   
  egcs-objc-970825.tar.gz               The objc runtime.


  egcs-testsuite-970825.tar.gz          The c & g++ testusites.
  egcs-g++-testsuite-970825.tar.gz      The g++ testsuite.


Note there is no patchkit to bring snapshot 970821 up to 970825.  Sorry.

We've actually only tested with the full snapshot; we're interested in
hearing if people can build the core or the core + particular language
add-ons.

There will likely be more layout changes in the future (sometime after the
first release).