build failure on x86_64 linux

[Wolfgang Pichl]

Hi,
I am trying to build gcl on x86_64 Linux (SLAMD64 distribution). I had any
kind of weird problems configuring and compiling the stable 2.6.7. After
applying 2 patches from Fedora and switching off brk randomization in
/proc/sys/kernel, I got at least 'configure' to work. 'Make' had a problem
linking libbfd.a, because the version I have requires to link with libz.
This could be solved by adding the linker flag -lz in makedefc and
makedefs, but finally compilation ended with a segmentation fault.
Therefore I decided to use the CVS. This worked much better, but finally
also failed.

The tool I am using:
gcc version: 4.2.4
binutils:    2.18.50
glibc:       2.7

These are my configure options:

./configure --libdir=/usr/local/lib64 --enable-readline --enable-ansi
--enable-xgcl --enable-tkconfig=/usr/lib64 --enable-tclconfig=/usr/lib64


Configure worked without any problem, however, after a long compilation
'make' ended up with the following error:

Pass1 signature discovery on 1095 functions ...
; (DEFUN PCL::|(FAST-METHOD SHARED-INITIALIZE :AFTER (STRUCTURE-CLASS T))|
         ...) is being compiled.
;; Warning: arg type mismatch in auto-proclamation (T T STRUCTURE-CLASS
                                                     T T) -> (T T T T T)

; (DEFUN PCL::|(FAST-METHOD SHARED-INITIALIZE :AFTER (STD-CLASS T))|
          ...) is being compiled.
;; Warning: arg type mismatch in auto-proclamation (T T PCL::STD-CLASS
                                                     T T) -> (T T T T T)

Error:
Fast links are on: do (si::use-fast-links nil) for debugging
Signalled by LET.
SIMPLE-ERROR: can't compile (PCL::FAST-METHOD PRINT-OBJECT (CONDITION T))

Broken at LET.  Type :H for Help.
1 (Continue) Retry compiling (CONDITIONS::|(PCL::FAST-METHOD PRINT-OBJECT
(CONDITION T))|
                                   NIL).
2 Return to top level.
COMMON-LISP-USER>>make[1]: *** [final] Error 255
rm recompile.h pre_init.lsp post_init.lsp
make[1]: Leaving directory `/usr/local/src/gcl/unixport'
make: *** [system] Error 2

Can anybody help me?

N.B. I posted this already yesterday, but obviously too early (after
subscribing, but before receiving the confirmation mail). Obviously my
first post has been ignored (as it should). Anyway, please excuse, if this
should appear doubly on the list.

Best regards
Wolfgang Pichl





_______________________________________________
Gcl-devel mailing list
Gcl-devel@...
http://lists.gnu.org/mailman/listinfo/gcl-devel

[Jerry James-3]

Hi Wolfgang,

On Tue, Jun 23, 2009 at 10:33 AM, Wolfgang
Pichl<wolfgang.pichl@...> wrote:
> Hi,
> I am trying to build gcl on x86_64 Linux (SLAMD64 distribution). I had any
> kind of weird problems configuring and compiling the stable 2.6.7. After
> applying 2 patches from Fedora and switching off brk randomization in
> /proc/sys/kernel, I got at least 'configure' to work. 'Make' had a problem
> linking libbfd.a, because the version I have requires to link with libz.
> This could be solved by adding the linker flag -lz in makedefc and makedefs,
> but finally compilation ended with a segmentation fault.

I'm the Fedora maintainer, so those patches are my fault. :-)  One of
them makes it unnecessary to switch off brk randomization for the
entire box, if that matters to you (gcl-2.6.8-unrandomize.patch).

Also, note that the -lz problem has been fixed in CVS, and a number of
the Fedora patches have been applied as well.

> Therefore I decided to use the CVS. This worked much better, but finally
> also failed.

That's a strange failure.  I'm going to take a wild shot in the dark
here.  Does adding -fno-strict-aliasing to CFLAGS help?
--
Jerry James
http://www.jamezone.org/


_______________________________________________
Gcl-devel mailing list
Gcl-devel@...
http://lists.gnu.org/mailman/listinfo/gcl-devel

[Wolfgang Pichl]

Hi Jerry,

On Tue, 23 Jun 2009, Jerry James wrote:

> Hi Wolfgang,
>
> I'm the Fedora maintainer, so those patches are my fault. :-)  One of
> them makes it unnecessary to switch off brk randomization for the
> entire box, if that matters to you (gcl-2.6.8-unrandomize.patch).

I tried two patches from a gcl-2.6.7-14.fc7 source RPM I happened two find
somewhere on the web. The first one, gcl-bash.patch, solved a syntax error
in the configure script, the other one, gcl-pers.patch, seems to be
supposed to work around the brk randomization problem, but this did not
work for me, I still had to disable the randomization. Probably there are
newer and better patches, I just used, what I found by googling. Anyway,
I'll give up with 2.6.7 and hope to get the CVS version to work.

>
> Also, note that the -lz problem has been fixed in CVS, and a number of
> the Fedora patches have been applied as well.

Yes, in the CVS the -lz problem and also the brk problem seem to be
fixed.
>
>> Therefore I decided to use the CVS. This worked much better, but finally
>> also failed.
>
> That's a strange failure.  I'm going to take a wild shot in the dark
> here.  Does adding -fno-strict-aliasing to CFLAGS help?
I'll try this tomorrow, because I am not in my office yet. Many thanks for
the tip.

best regards
Wolfgang Pichl



_______________________________________________
Gcl-devel mailing list
Gcl-devel@...
http://lists.gnu.org/mailman/listinfo/gcl-devel

[Jerry James-3]

On Tue, Jun 23, 2009 at 2:36 PM, Wolfgang
Pichl<wolfgang.pichl@...> wrote:
> I tried two patches from a gcl-2.6.7-14.fc7 source RPM I happened two find
> somewhere on the web. The first one, gcl-bash.patch, solved a syntax error
> in the configure script, the other one, gcl-pers.patch, seems to be supposed
> to work around the brk randomization problem, but this did not work for me,
> I still had to disable the randomization. Probably there are newer and
> better patches, I just used, what I found by googling. Anyway, I'll give up
> with 2.6.7 and hope to get the CVS version to work.

Oh, so I lied then.  I'm not responsible for those patches. :-)  I
stashed mine here: http://jjames.fedorapeople.org/gcl/ (but, like I
said, note that some of those have already been applied to CVS).

> I'll try this tomorrow, because I am not in my office yet. Many thanks for
> the tip.

I hope it does you some good.  Let us know if it helps or not.
--
Jerry James
http://www.jamezone.org/


_______________________________________________
Gcl-devel mailing list
Gcl-devel@...
http://lists.gnu.org/mailman/listinfo/gcl-devel

[Wolfgang Pichl]

Hi,
On Tue, 23 Jun 2009, Jerry James wrote:

> Hi Wolfgang,
>
>
> That's a strange failure.  I'm going to take a wild shot in the dark
> here.  Does adding -fno-strict-aliasing to CFLAGS help?
> --
> Jerry James
> http://www.jamezone.org/
>
>

Unfortunately it does not help. Exactly the same error.
Could this be a 64bit issue? I'll try to compile the CVS on my older 32bit
box running Slackware-12.2.
I really don't understand this error message. seems to be something
internal to Lisp. The error seems to occur during compiling or testing
some Lisp modules with the preliminarily compiled glc...

regards
Wolfgang



_______________________________________________
Gcl-devel mailing list
Gcl-devel@...
http://lists.gnu.org/mailman/listinfo/gcl-devel

[Camm Maguire-2]

Greetings, and thanks so much for your report!

I'd like to chae down the cvs issue, but to get going quickly I'd like
to make sure you are using the 2.6.8pre cvs branch.

My apologies for this somewhat sscatterred state of affairs, but here
is the GCL development convention:

x.y.z, y even -- stable series
       y odd  -- development series

2.6.8 will be the next stable release.  It really should have been
released some time ago.  2.7.0, aka cvs HEAD, is the development
snapshot and can experience breakages from time to time.

To get 2.6.8pre, which is the branch that will be tagged as 2.6.8 when
released, do

cvs -z9 -q -d:pserver:anonymous@...:/sources/gcl \
     co -d gcl-2.6.8pre -r Version_2_6_8pre gcl

To get cvs head, do

cvs -z9 -q -d:pserver:anonymous@...:/sources/gcl \
     co -d gclcvs-2.7.0 gcl

Finally, the Debian packages labeled 2.6.7-??? are actually 2.6.8pre
snapshots.  This is to ensure that 2.6.8 will compile on all
architectures supporting all applications before final release.

Please let me know if problems persist.

Take care,

Wolfgang Pichl <wolfgang.pichl@...> writes:

> Hi,
> I am trying to build gcl on x86_64 Linux (SLAMD64 distribution). I had
> any kind of weird problems configuring and compiling the stable
> 2.6.7. After applying 2 patches from Fedora and switching off brk
> randomization in /proc/sys/kernel, I got at least 'configure' to
> work. 'Make' had a problem linking libbfd.a, because the version I
> have requires to link with libz. This could be solved by adding the
> linker flag -lz in makedefc and makedefs, but finally compilation
> ended with a segmentation fault.
> Therefore I decided to use the CVS. This worked much better, but
> finally also failed.
>
> The tool I am using:
> gcc version: 4.2.4
> binutils:    2.18.50
> glibc:       2.7
>
> These are my configure options:
>
> ./configure --libdir=/usr/local/lib64 --enable-readline --enable-ansi
> --enable-xgcl --enable-tkconfig=/usr/lib64
> --enable-tclconfig=/usr/lib64
>
>
> Configure worked without any problem, however, after a long
> compilation 'make' ended up with the following error:
>
> Pass1 signature discovery on 1095 functions ...
> ; (DEFUN PCL::|(FAST-METHOD SHARED-INITIALIZE :AFTER (STRUCTURE-CLASS T))|
>         ...) is being compiled.
> ;; Warning: arg type mismatch in auto-proclamation (T T STRUCTURE-CLASS
>                                                     T T) -> (T T T T T)
>
> ; (DEFUN PCL::|(FAST-METHOD SHARED-INITIALIZE :AFTER (STD-CLASS T))|
>          ...) is being compiled.
> ;; Warning: arg type mismatch in auto-proclamation (T T PCL::STD-CLASS
>                                                     T T) -> (T T T T T)
>
> Error:
> Fast links are on: do (si::use-fast-links nil) for debugging
> Signalled by LET.
> SIMPLE-ERROR: can't compile (PCL::FAST-METHOD PRINT-OBJECT (CONDITION T))
>
> Broken at LET.  Type :H for Help.
> 1 (Continue) Retry compiling (CONDITIONS::|(PCL::FAST-METHOD
> PRINT-OBJECT (CONDITION T))|
>                                   NIL).
> 2 Return to top level.
> COMMON-LISP-USER>>make[1]: *** [final] Error 255
> rm recompile.h pre_init.lsp post_init.lsp
> make[1]: Leaving directory `/usr/local/src/gcl/unixport'
> make: *** [system] Error 2
>
> Can anybody help me?
>
> N.B. I posted this already yesterday, but obviously too early (after
> subscribing, but before receiving the confirmation mail). Obviously my
> first post has been ignored (as it should). Anyway, please excuse, if
> this
> should appear doubly on the list.
>
> Best regards
> Wolfgang Pichl
>
>
>
>
>
> _______________________________________________
> Gcl-devel mailing list
> Gcl-devel@...
> http://lists.gnu.org/mailman/listinfo/gcl-devel
>
>
>
>

--
Camm Maguire        camm@...
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah


_______________________________________________
Gcl-devel mailing list
Gcl-devel@...
http://lists.gnu.org/mailman/listinfo/gcl-devel

[Wolfgang Pichl]

Hi,

On Thu, 25 Jun 2009, Camm Maguire wrote:

>
> My apologies for this somewhat sscatterred state of affairs, but here
> is the GCL development convention:
>
> x.y.z, y even -- stable series
>       y odd  -- development series
>
> 2.6.8 will be the next stable release.  It really should have been
> released some time ago.  2.7.0, aka cvs HEAD, is the development
> snapshot and can experience breakages from time to time.
>
> To get 2.6.8pre, which is the branch that will be tagged as 2.6.8 when
> released, do
>
> cvs -z9 -q -d:pserver:anonymous@...:/sources/gcl \
>     co -d gcl-2.6.8pre -r Version_2_6_8pre gcl
>
> To get cvs head, do
>
> cvs -z9 -q -d:pserver:anonymous@...:/sources/gcl \
>     co -d gclcvs-2.7.0 gcl

Obviously I have the cvs head. I simply did

cvs -z9 -q -d:pserver:anonymous@...:/sources/gcl \
      co gcl

without specifying any version. I was not aware of different versions
available; browsing the cvs repository only shows one single version. My
apologies, I am not very experienced with cvs...
Obviously I should try 2.6.8pre. I'll do that now and report success or
failure soon.

Many thanks
Wolfgang Pichl



_______________________________________________
Gcl-devel mailing list
Gcl-devel@...
http://lists.gnu.org/mailman/listinfo/gcl-devel

[Wolfgang Pichl]

Hi,
compilation of gcl-2.6.8pre works out of the box. I built both a 64bit
version under Slamd64-12.2 and a 32bit version under Slackware-12.2.
Everything fine. So the compilation problem was due to the fact that I
used the cvs head.

I saw that in 2.6.8pre there is still a small problem building the info
documentation. It is unimportant for me, however, I think I should report
it. At the end of the compilation I get this:

/usr/local/src/gcl-2.6.8pre/info/gcl-si/index.html: Is a directory
make[1]: [gcl-si/index.html] Error 1 (ignored)
mkdir -p gcl-tk/index.html
touch gcl-tk/index.html
makeinfo --html  gcl-tk.texi
/usr/local/src/gcl-2.6.8pre/info/gcl-tk/index.html: Is a directory
make[1]: [gcl-tk/index.html] Error 1 (ignored)
make[1]: Leaving directory `/usr/local/src/gcl-2.6.8pre/info'

best regards, and many thanks for the help
Wolfgang Pichl



_______________________________________________
Gcl-devel mailing list
Gcl-devel@...
http://lists.gnu.org/mailman/listinfo/gcl-devel