[tree-ssa, cvs] spec failure: unrecognized spec option '<'

[tree-ssa, cvs] spec failure: unrecognized spec option '<'

[Melvin Hadasht]

[-- Attachment #1: Type: text/plain, Size: 1663 bytes --]

Hi,

While compiling tree-ssa from CVS, I get the following error:

gcc: spec failure: unrecognized spec option '<'
cc1: error: output filename specified twice
make[1]: *** [g++spec.o] Error 1
make[1]: Leaving directory `/usr/local/src/cvs/objdir/gcc'
make: *** [all-gcc] Error 2

(the command that issues this is listed in the attachement)

I'm compiling on PII, linux 2.6.0test9, gcc-3.3.2 (compiled from source). I compiled in a separate directory
and configured with --enable-languages=c,c++,fortran (although even withou c++ I hit the same specs problem).

The gcc/specs file contains the following section, note the second last line:
*cc1_cpu:
%{!mtune*: %{m386:mtune=i386 %n`-m386' is deprecated. Use `-march=i386' or `-mtune=i386' instead.
} %{m486:-mtune=i486 %n`-m486' is deprecated. Use `-march=i486' or `-mtune=i486' instead.
} %{mpentium:-mtune=pentium %n`-mpentium' is deprecated. Use `-march=pentium' or `-mtune=pentium' instead.
} %{mpentiumpro:-mtune=pentiumpro %n`-mpentiumpro' is deprecated. Use `-march=pentiumpro' or `-mtune=pentiumpro' instead.
} %{mcpu=*:-mtune=%* %n`-mcpu=' is deprecated. Use `-mtune=' or '-march=' instead.
}} %<mcpu=* %{mintel-syntax:-masm=intel %n`-mintel-syntax' is deprecated. Use `-masm=intel' instead.
} %{mno-intel-syntax:-masm=att %n`-mno-intel-syntax' is deprecated. Use `-masm=att' instead.
}

if I remove "%<mcpu=*", the compilation can be resumed. I'm doing this since many weeks to be able to build tree-ssa.
I don't remember if I did the same for gcc-3.3.2 (as removing this has become natural for me). 

Does someone know what could be the origin of this problem?  

Thanks in advance.

-- 
Melvin Hadasht

[-- Attachment #2: gcc-error-spec --]
[-- Type: text/plain, Size: 1707 bytes --]

(SHLIB_LINK=' ./xgcc -B./ -B/usr/local/i686-pc-linux-gnu/bin/ -isystem /usr/local/i686-pc-linux-gnu/include -isystem /usr/local/i686-pc-linux-gnu/sys-include -L/usr/local/src/cvs/objdir/gcc/../ld -O2  -DIN_GCC    -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -isystem ./include  -fPIC -g -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -shared -nodefaultlibs -Wl,--soname=@shlib_so_name@.so.1 -Wl,--version-script=@shlib_map_file@ -o @shlib_dir@@shlib_so_name@.so.1 @multilib_flags@ @shlib_objs@ -lc && rm -f @shlib_base_name@.so && ln -s @shlib_dir@@shlib_so_name@.so.1 @shlib_base_name@.so' \
SHLIB_MULTILIB=''; \
gcc -c   -g -DIN_GCC   -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -pedantic -Wno-long-long -fno-common   -DHAVE_CONFIG_H    -DSTANDARD_STARTFILE_PREFIX=\"../../../\" -DSTANDARD_EXEC_PREFIX=\"/usr/local/lib/gcc/\" -DSTANDARD_LIBEXEC_PREFIX=\"/usr/local/libexec/gcc/\" -DDEFAULT_TARGET_VERSION=\"3.5-tree-ssa\" -DDEFAULT_TARGET_MACHINE=\"i686-pc-linux-gnu\" -DSTANDARD_BINDIR_PREFIX=\"/usr/local/bin/\" -DTOOLDIR_BASE_PREFIX=\"../../../../\" `test "X$${SHLIB_LINK}" = "X" || test "yes" != "yes" || echo "-DENABLE_SHARED_LIBGCC"` `test "X${SHLIB_MULTILIB}" = "X" || echo "-DNO_SHARED_LIBGCC_MULTILIB"` \
        -I. -I. -I/usr/local/src/cvs/gcc/gcc -I/usr/local/src/cvs/gcc/gcc/. -I/usr/local/src/cvs/gcc/gcc/../include -I/usr/local/src/cvs/gcc/gcc/../libbanshee/libcompat -I/usr/local/src/cvs/gcc/gcc/../libbanshee -I/usr/local/src/cvs/gcc/gcc/../libbanshee/points-to  /usr/local/src/cvs/gcc/gcc/cp/g++spec.c)
gcc: spec failure: unrecognized spec option '<'
cc1: error: output filename specified twice
make: *** [g++spec.o] Error 1

Re: [tree-ssa, cvs] spec failure: unrecognized spec option '<'

[Dara Hazeghi]

Hello,

looks like you found a bug. For me though, I only see
it by visual inspection, ie I don't get any error
building. Still, it looks like a bug in either case.
File a report through bugzilla (or I will, if I get a
chance). Thanks,

Dara

__________________________________
Do you Yahoo!?
Protect your identity with Yahoo! Mail AddressGuard
http://antispam.yahoo.com/whatsnewfree

Re: [tree-ssa, cvs] spec failure: unrecognized spec option '<'

[Melvin Hadasht]

On Sat Nov 08, at 11:14, Dara wrote:

> looks like you found a bug. For me though, I only see
> it by visual inspection, ie I don't get any error
> building. Still, it looks like a bug in either case.
> File a report through bugzilla (or I will, if I get a
> chance). Thanks,

I'll try to recompile gcc-3.3.2 to see if I have the same problem before
submitting a bug report.

Thanks.

-- 
Melvin Hadasht

Re: [tree-ssa, cvs] spec failure: unrecognized spec option '<'

[Dara Hazeghi]

Hello,

I just checked on mainline cvs, and the same problem's
there. However on 3.3 branch it looks fine. I'm
guessing this was introduced when Neil rewrotes
specs/options handling on mainline.

Dara

P.S. I'm still puzzled with the error you get though
(I see nothing)...

__________________________________
Do you Yahoo!?
Protect your identity with Yahoo! Mail AddressGuard
http://antispam.yahoo.com/whatsnewfree

Re: [tree-ssa, cvs] spec failure: unrecognized spec option'<'

[Dara Hazeghi]

Hello

> Ok, thanks for checking: on my pII 266,
> compilation is not instantaneous.

Yeah, gcc compile time isn't exactly spectacular :-)

>Reading specs from ./specs
>Configured with:
/usr/local/src/cvs/gcc-3.3.2/>configure 
>Thread model: posix
>gcc driver version 3.3.2 executing gcc version
>3.5-tree-ssa

See this is what's weird, and symptomatic of a broken
installation. For some reason, your 3.3.2 driver is
running the 3.5 compiler. I'm guessing the 3.3.x
driver is less forgiving about spec file typos,
hence...

>gcc: spec failure: unrecognized spec option '<'

If you go to the prefix where you installed 3.3.x, and
do a find on cc1, what do you get? More importantly,
if you run that cc1, what version does it say it is?
Also, what is you $PATH?

>Concerning the ./specs file, removing the >"%<mcpu=*"
is not sufficient, 
>a closing brace "}" must also be adjusted.

Right.

>I don't have any idea on how to find where the
>problem is, but I'm willing to do any tests. 
>Should I file a bug report?

You can certainly file a bug report, mentioning the
typo in specs, and give the particulars. The other
error your seeing is more likely part of a screwed up
installation (could also be a bug...). 

If you want to find what's causing the bug, you can
try to figure out what makefile rule generates
gcc/specs (in your build directory), and so figuree
out where the original typo comes from. I'll also take
a look this afternoon and see what I can come up
with...

Cheers,

Dara

P.S. Please cc: replies, since I'm not subscribed to gcc-help@...

__________________________________
Do you Yahoo!?
Protect your identity with Yahoo! Mail AddressGuard
http://antispam.yahoo.com/whatsnewfree

Re: [tree-ssa, cvs] spec failure: unrecognized spec option'<'

[Melvin Hadasht]

Hi,

I did some more tests:
Configuring with --enable-languages=c (without c++ or f95), and doing
a make (no bootstrap) compiles correctly (but the specs did have the bug).
 I'm currently doing a bootstrap to test. (I'm still using tree-ssa from
cvs 2003-11-07 13:20 UTC)

>>Reading specs from ./specs
>>Configured with:
>>/usr/local/src/cvs/gcc-3.3.2/>configure 
>>Thread model: posix
>>gcc driver version 3.3.2 executing gcc version
>>3.5-tree-ssa

>See this is what's weird, and symptomatic of a broken
>installation. For some reason, your 3.3.2 driver is
>running the 3.5 compiler. 

Yes, it reads the ./specs it has found. According to the command line
that triggered the error (file attached in my original post), I don't see
why it reads the specs from the current directory.

Testing just gcc -v in a directory containing an empty specs shows
that gcc (3.3.2) reads the specs from that file. So it seems that gcc 
searches for the specs in the current directory first. Is this expected? 
If yes, then that would explain the above. And 3.3.2 being more
strict in reading the specs, it will also explain why I have the error.
So I assume you do not use 3.3.2 to compile tree-ssa?

>I'm guessing the 3.3.x
>driver is less forgiving about spec file typos,
>hence...


>>gcc: spec failure: unrecognized spec option '<'

>If you go to the prefix where you installed 3.3.x, and
>do a find on cc1, what do you get? 
cd /usr/local
find -name 'cc1'
./lib/gcc-lib/i686-pc-linux-gnu/2.95.4/cc1
./lib/gcc-lib/i686-pc-linux-gnu/3.3.2/cc1

>More importantly,
>if you run that cc1, what version does it say it is?
>Also, what is you $PATH?

/usr/local/lib/gcc-lib/i686-pc-linux-gnu/3.3.2/cc1 --version
GNU C version 3.3.2 (i686-pc-linux-gnu)
        compiled by GNU C version 3.3.2.
GGC heuristics: --param ggc-min-expand=38 --param ggc-min-heapsize=15861


which gcc
/usr/local/bin/gcc

/usr/local/bin/gcc -v
Reading specs from /usr/local/lib/gcc-lib/i686-pc-linux-gnu/3.3.2/specs
Configured with: /hannibal/src/cvs/gcc-3.3.2/configure 
Thread model: posix
gcc version 3.3.2

echo $PATH
/usr/local/qt/bin:/home/mhadasht/bin:/usr/local/bin:/usr/bin:/usr/X11R6/bin:/bin:/usr/local/nmh/bin:/usr/games/bin:/usr/games:.:/usr/local/teTeX/bin:/usr/local/MuPAD20/share/bin:/usr/local/mysql/bin

> You can certainly file a bug report, mentioning the
> typo in specs, and give the particulars. The other
> error your seeing is more likely part of a screwed up
> installation (could also be a bug...). 

Ok, I'll do some other tests and I'll file a bug report.

> If you want to find what's causing the bug, you can
> try to figure out what makefile rule generates
> gcc/specs (in your build directory), and so figuree
> out where the original typo comes from.

I'll try that.

>  I'll also take
> a look this afternoon and see what I can come up
> with...

Ok, thanks for you help.

Cheers,

Melvin

> P.S. Please cc: replies, since I'm not subscribed to gcc-help@...
Sorry. Done!

Re: [tree-ssa, cvs] spec failure: unrecognized spec option'<'

[Dara Hazeghi]

Hello,

> I did some more tests:
> Configuring with --enable-languages=c (without c++
> or f95), and doing
> a make (no bootstrap) compiles correctly (but the
> specs did have the bug).
>  I'm currently doing a bootstrap to test. (I'm still
> using tree-ssa from
> cvs 2003-11-07 13:20 UTC)

So here's a question. If you can get it to the point
where it fails again, would you mind running that
final failing commandline, but gradually trimming off
options, and experimenting, to see what it can be
reduced to? Hopefully this won't require a full
bootstrap...

I think this approach is most likely to find out why
the 3.3 driver is invoking the newly built 3.5
compiler.

Oh, as for the typo, I figured out where it was coming
from, and sent a message to the main gcc list.
Hopefully somebody will apply a patch for it soon (if
not I'll open a bug report to remind them).

Cheers,

Dara

__________________________________
Do you Yahoo!?
Protect your identity with Yahoo! Mail AddressGuard
http://antispam.yahoo.com/whatsnewfree

Re: [tree-ssa, cvs] spec failure: unrecognized spec option'<'

[Melvin Hadasht]

Hi Dara,

> Oh, as for the typo, I figured out where it was coming
> from, and sent a message to the main gcc list.
> Hopefully somebody will apply a patch for it soon (if
> not I'll open a bug report to remind them).

Ok, I read the thread on gcc. The threads concludes that
there is no bug in the specs and the "<" thing is expected.
The problem is that my gcc-3.3.2 reads the specs in the 
current directory, which is not expected.
So it looks definitely as a problem
with my gcc-3.3.2, or my build environment. I'll try to install
another gcc from rpms and not from source (that is new to me :) ).
So I won't post a bug report, as there is no bug.

Thank you very much for your help.

Best regards,

Melvin

-- 
Melvin Hadasht

Re: [tree-ssa, cvs] spec failure: unrecognized spec option'<'

[Dara Hazeghi]

--- Melvin Hadasht <melvin.hadasht@free.fr> wrote:
Hello,

> Ok, I read the thread on gcc. The threads 
> concludes that there is no bug in the specs and >
the "<" thing is expected. The problem is that 
> my gcc-3.3.2 reads the specs in the  current 
> directory, which is not expected.

Yeah, that seems to be the situation. Good luck
installing from rpms. If you have any more problems,
by all means report back though. GCC may not be
trivial to install, but we try not to make it too
impossible!

Dara

__________________________________
Do you Yahoo!?
Protect your identity with Yahoo! Mail AddressGuard
http://antispam.yahoo.com/whatsnewfree