Bootstrap failure in toplev.c

Bootstrap failure in toplev.c

[Gerald Pfeifer]

The following patch seems to have broken bootstrap (seen on
i386-unknown-freebsd4.7):

2003-02-18  Matt Austern <austern@apple.com>

        * langhooks.h, langhooks-def.h: introduce new langhook,
        final_write_globals, with write_global_declarations as default.
        * toplev.c: Move invocation of wrapup_global_declarations from
        compile_file to new function, write_global_declarations.  Change
        compile_file to use final_write_globals hook.  Change
        wrapup_global_declarations so writing to DECL_DEFER_OUTPUT is
        conditional.

stage1/xgcc -Bstage1/ -B/sw/gcc-current/i386-unknown-freebsd4.7/bin/ -c   -g -O2 -DIN_GCC   -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -Wtraditional -pedantic -Wno-long-long  -Werror   -DHAVE_CONFIG_H    -I. -I. -I/sw/test/gcc/cvs/gcc -I/sw/test/gcc/cvs/gcc/. -I/sw/test/gcc/cvs/gcc/config -I/sw/test/gcc/cvs/gcc/../include /sw/test/gcc/cvs/gcc/tree.c -o tree.o
/sw/test/gcc/cvs/gcc/toplev.c:2230: warning: function declaration isn't a prototype
gmake[2]: *** [toplev.o] Error 1
gmake[2]: *** Waiting for unfinished jobs....
gmake[2]: Leaving directory `/files/pfeifer/OBJ-0218-2108/gcc'
gmake[1]: *** [stage2_build] Error 2
gmake[1]: Leaving directory `/files/pfeifer/OBJ-0218-2108/gcc'
gmake: *** [bootstrap-lean] Error 2

Did this patch really pass a full bootstrap in a recent gcc.gnu.org tree?

Gerald
-- 
Gerald "Jerry"   pfeifer@dbai.tuwien.ac.at   http://www.pfeifer.com/gerald/

Re: Bootstrap failure in toplev.c

[Matt Austern]

On Tuesday, February 18, 2003, at 12:20 PM, Gerald Pfeifer wrote:

> The following patch seems to have broken bootstrap (seen on
> i386-unknown-freebsd4.7):

I'll take a look.

> Did this patch really pass a full bootstrap in a recent gcc.gnu.org 
> tree?

Depends on what you mean by recent.  It passed a bootstrap when I
submitted it, which is about a week ago.  Perhaps there's a bad
interaction between this patch and something that happened between
submission and approval.  More later...

			--Matt

Re: Bootstrap failure in toplev.c

[Graham Stott]

Gerald Pfeifer wrote:
> The following patch seems to have broken bootstrap (seen on
> i386-unknown-freebsd4.7):
Same here on i686-pc-linux-gnu

> 
> Did this patch really pass a full bootstrap in a recent gcc.gnu.org tree?
> 
I doubt it :-(

> Gerald

I think the implementation of write_global_declarations should be
moved to langhooks.c where the other default lang hooks are.

Graham
ps. The routine is also incorrectly formatted the return type should be
on it's own line.

Re: Bootstrap failure in toplev.c

[Matt Austern]

On Tuesday, February 18, 2003, at 12:20 PM, Gerald Pfeifer wrote:

> The following patch seems to have broken bootstrap (seen on
> i386-unknown-freebsd4.7):

OK, now I need help.  I just did the following things on a Linux system
(configuration i686-pc-linux-gnu; it's an RH 7.2 installation, with 
their
gcc  2.96-112.7.2 installed as the system compiler.)

  - cd gcc34     # this is mainline; I verified there are no sticky tags
  - cvs update
  - cd ..
  - mkdir obj34
  - cd obj34
  - ../gcc34/configure --prefix=/home/austern/root
  - make
  - make install

The compiler bootstrapped and installed without error.  What did you
do differently? At what point in the process did you see a failure?

I'm entirely willing to believe that I goofed and that something breaks
under some specific circumstances, but I don't yet know what those
circumstances are.

			--Matt

Re: Bootstrap failure in toplev.c

[Richard Henderson]

On Tue, Feb 18, 2003 at 02:40:25PM -0800, Matt Austern wrote:
>  - make
> 
> The compiler bootstrapped and installed without error.

This is not a bootstrap.  "make bootstrap" is a bootstrap.



r~

Re: Bootstrap failure in toplev.c

[Steven Bosscher]

Op di 18-02-2003, om 23:40 schreef Matt Austern:
> 
>   - cd gcc34     # this is mainline; I verified there are no sticky tags
>   - cvs update
>   - cd ..
>   - mkdir obj34
>   - cd obj34
>   - ../gcc34/configure --prefix=/home/austern/root
>   - make

make bootstrap??

>   - make install
> 
> The compiler bootstrapped and installed without error.  What did you
> do differently? At what point in the process did you see a failure?
> 
> I'm entirely willing to believe that I goofed and that something breaks
> under some specific circumstances, but I don't yet know what those
> circumstances are.
> 

Greetz
Steven

Re: Bootstrap failure in toplev.c

[Matt Austern]

On Tuesday, February 18, 2003, at 02:43 PM, Steven Bosscher wrote:

> Op di 18-02-2003, om 23:40 schreef Matt Austern:
>>
>>   - cd gcc34     # this is mainline; I verified there are no sticky 
>> tags
>>   - cvs update
>>   - cd ..
>>   - mkdir obj34
>>   - cd obj34
>>   - ../gcc34/configure --prefix=/home/austern/root
>>   - make
>
> make bootstrap??

Oops!  Yes.  OK, fix coming up.

			--Matt

Re: Bootstrap failure in toplev.c

[Loren James Rittle]

In article <F8CA77D7-4391-11D7-9473-00039390D9E0@apple.com> Matt writes:

>> The following patch seems to have broken bootstrap (seen on
>> i386-unknown-freebsd4.7):

> OK, now I need help.

Yes, you didn't get to experience the full pleasure of -Werror
checking with your technique.  As RTH mentioned, you are not doing a
full bootstrap.  For the record, you also skipped the required 'make
check' (a step which is NOT optional for anything other than trivial
bug fixes where trivial is in the eye of the beholder).  Here is your
technique recast to a more correct form:

>  - cd gcc34     # this is mainline; I verified there are no sticky tags
>  - cvs update

# You might be subtly testing a different set of sources from all
# other developers, if you skip this step:

./contrib/gcc_update --touch

# If you do not trust running that script, consider that you are
# trusting all shell script fragments in all gcc Makefiles.
# Either way, feel free to inspect it before running it. ;-)

>  - cd ..
>  - mkdir obj34
>  - cd obj34
>  - ../gcc34/configure --prefix=/home/austern/root

make bootstrap

make -k check

# Look at regression report versus baseline for target.  There are
# many posted daily for common targets but it is best to have one for
# your exact system since results may vary slightly due to binutils
# version, etc.

>  - make install

# Safe to commit/post here.

Regards,
Loren