build error of fricas-1.0.7
Bertfried Fauser
I have frequently tried to build fricas (1.0.7) and failed for reasons
I cannot detect myself. Any help what I do wrong is greatly
appreciated.
I did the following:
* I downloaded 'fricas-1.0.7-full.tar' (unbziped) from sourceforge.
* I have installed sbcl (1.0.30) and gcl-2.6 (compiles my AXIOM)
* I have now notangle, noweb in my path (configure finds them)
* configure seems to look for util.h which seems to be not on my
machine, but doesn't complein about it.
* I run confugure as
> ./configure --with-lisp=[sbcl|gcl]
and it doesn't complain on anything, so I go ahead with
* make
which results (for sbcl) in the following last lines of make output:
Added 56 pages and/or macros from ZDSOLVE.pht
Added 16 pages and/or macros from ZLINDEP.pht
make[2]: Leaving directory `/home/fauser/fricas-1.0.7/src/paste'
cd aldor && make
make[2]: Entering directory `/home/fauser/fricas-1.0.7/src/aldor'
Makefile:262: domains.mk: No such file or directory
echo "domains := \\" > domains.mk
grep '^)abbrev' /home/fauser/fricas-1.0.7/src/algebra/*.spad | gawk
'{print $3,"\\"}' >> domains.mk
echo >> domains.mk
make[2]: Leaving directory `/home/fauser/fricas-1.0.7/src/aldor'
make[2]: Entering directory `/home/fauser/fricas-1.0.7/src/aldor'
svn cat https://svn.origo.ethz.ch/algebraist/trunk/aldor/lib/libax0/axiom.as
> axiom.as
svn: PROPFIND request failed on '/algebraist/trunk/aldor/lib/libax0/
axiom.as'
svn: PROPFIND of '/algebraist/trunk/aldor/lib/libax0/axiom.as': could
not connect to server (https://svn.origo.ethz.ch)
make[2]: *** [axiom.as] Error 1
make[2]: Leaving directory `/home/fauser/fricas-1.0.7/src/aldor'
make[1]: *** [all-aldor] Error 2
make[1]: Leaving directory `/home/fauser/fricas-1.0.7/src'
make: *** [all-src] Error 2
fauser@vis5:~/fricas-1.0.7>
by the way, I run a SuSE 10.1 linux:
Linux vis5.pc 2.6.16.21-0.25-smp #1 SMP Tue Sep 19 07:26:15 UTC 2006
i686 i686 i386 GNU/Linux
with gcc (GCC) 4.1.0 (SUSE Linux)
It somehow seems I cannot get aldor files, can I compile without these
files? Do I mandatorily need to compile with aldor support? I found no
switch to prevent this (but I do not really know where to search)
I am behind a proxi which doesn't allow to make svn
connections......
Kind regards
BF.
Waldek Hebisch
The switch '--enable-aldor=no' should disable aldor build.
Alternatively, doing
unset ALDORROOT
should prevent building Aldor. Concerning svn: it is possible
to build Aldor interface without svn, see INSTALL.aldor
(however, if you use release you will probably hit problem
due to empty domains.mk).
Ralf, I think that Aldor interface should be disabled by default,
and only build when explicitely requested.
--
Waldek Hebisch
heb...@math.uni.wroc.pl
Ralf Hemmecke
>> cd aldor && make
>> make[2]: Entering directory `/home/fauser/fricas-1.0.7/src/aldor'
>> Makefile:262: domains.mk: No such file or directory
>> echo "domains := \\" > domains.mk
>> grep '^)abbrev' /home/fauser/fricas-1.0.7/src/algebra/*.spad | gawk
>> '{print $3,"\\"}' >> domains.mk
>> echo >> domains.mk
> Ralf, I think that Aldor interface should be disabled by default,
> and only build when explicitely requested.
I don't think so. ;-)
That problem is related to
http://groups.google.com/group/fricas-devel/browse_thread/thread/19bb2ca0977f2f53/3825f3cacb4766c4#3825f3cacb4766c4
The reason is that compiling the source distribution, no .spad files
will be generated. (I have now not actually checked this, but most
probably you will not find .spad files in your build directory
/home/fauser/fricas-1.0.7/src/algebra/, right?)
I've constructed the interface with having an SVN checkout in mind...
This will be fixed hopefully within a week. I'm currently testing.
You should got to
/home/fauser/fricas-1.0.7/src/algebra/Makefile.in
and replace the target for domains.mk by the following:
domains.mk:
echo "domains := \\" > $@
echo ')lisp (dolist (c (|allConstructors|)) (format t "~A \\~%"
(|constructor?| c))) (quit)'|($(INTERPSYS))|grep -v '.*-'|grep -v
NIL|grep '\\' |sort >> $@
echo >> $@
Patch is attached. In fact, with this patch libaxiom.al includes 3 more
files, because there is a bug in the grep expression. (I should just
have searched for '^)abb' (no explicit r).
Then re-configure and start make again.
./configure --with-lisp=sbcl
make
Still problems?
BTW, why don't you use a scratch directory for your build? You could
keep your source directory uncluttered with all those generated files.
mkdir $HOME/fricas-build
cd $HOME/fricas-build
$HOME/fricas-1.0.7/configure --with-lisp=sbcl
--prefix=$HOME/fricas-install-dir
make
make install
In fact, libaxiom.al could be also distributed since it is machine
independent. That together with a 2-minute-build will be actually enough
to get the aldor-interface working. Unfortunately, libaxiom.al cannot
yet distributed together with FriCAS, since we still have no right to
include the few .as files from aldor.org under the Axiom (mBSD) license.
:-( Life would be so much easier...
Ralf
Ralf Hemmecke
> files? Do I mandatorily need to compile with aldor support?
No. Use the most obvious: configure --disable-aldor. ;-)
> I am behind a proxi which doesn't allow to make svn connections......
Hmm, I don't know which port SVN actually wants to access, but
https://svn.origo.ethz.ch/algebraist/trunk/aldor/lib/libax0/axiom.as
is an ordinary secured https URL. Hmm, maybe I should make 'wget' the
first option. Oh... why must we have this license problem? :-(
Could you try
make SVN=
this will try 'curl' if you have that or
make SVN= CURL=
this will try 'wget' (if you have that).
Otherwise get the files manually (see INSTALL.aldor).
Hope that helps.
Ralf
Waldek Hebisch
>
> > Ralf, I think that Aldor interface should be disabled by default,
> > and only build when explicitely requested.
>
> I don't think so. ;-)
>
> will be generated. (I have now not actually checked this, but most
> probably you will not find .spad files in your build directory
> /home/fauser/fricas-1.0.7/src/algebra/, right?)
I believe this part (once your patch is applied) will be fine.
But
>
> I've constructed the interface with having an SVN checkout in mind...
> This will be fixed hopefully within a week. I'm currently testing.
>
I am not sure what you want to say above. I think that running
SVN checkout for unsuspecting user is bad. I agreed to this
because it looked very unlikely that Aldor interface build will
be triggered for users who did not intend to build the interface.
But unfortunatly recent cases show that this happens.
Let me say this in different way:
I think that build process should do network operations only
after apropriate explicit user action. IMHO '--enable-aldor'
may qualify as beeing explicit enough. Just building
FriCAS is not.
--
Waldek Hebisch
heb...@math.uni.wroc.pl
Ralf Hemmecke
>> I've constructed the interface with having an SVN checkout in mind...
>> This will be fixed hopefully within a week. I'm currently testing.
> I am not sure what you want to say above.
Ooops, you misinterpreted that. What I meant was. That for building the
aldor interface I assumed that fricas was build from an SVN checkout,
i.e. the .spad files would be there.
> I think that running
> SVN checkout for unsuspecting user is bad.
I know and I agree. Just last night I asked Stephen Watt and Mike Dewar
to release the respective .as files under a license that allows us to
include them into the FriCAS repository. No answer yet.
> I agreed to this
> because it looked very unlikely that Aldor interface build will
> be triggered for users who did not intend to build the interface.
> But unfortunatly recent cases show that this happens.
>
> Let me say this in different way:
>
> I think that build process should do network operations only
> after apropriate explicit user action. IMHO '--enable-aldor'
> may qualify as beeing explicit enough. Just building
> FriCAS is not.
Waldek, if you think it's wiser to ask the user to explicitly give the
--enable-aldor switch at configure time, I don't oppose too much. I
think the situation is different now.
Furthermore, William Stein allowed me to distribute a fricasaldor.spkg
(which would basically just contain libaxiom.al). So when we generate a
new fricas release, we could upload an appropriate .spkg and also ask
users (or have a script) to fetch that spkg from
http://sagemath.org/packages/optional/. They would thus get libaxiom.al
from there and can (without much build time) complete the necessary
compilation of just the runtime files locally.
Ralf
PS: Just give me a few more days to finish the infrastructure stuff.
leh...@bayou.uni-linz.ac.at
> Ralf, I think that Aldor interface should be disabled by default,
> and only build when explicitely requested.
Currently the aldor interface is enabled only if aldor is present.
This indicates that the user is somehow interested in aldor
and might want to have it available in fricas, too.
Wouldn't it be possible to fetch the missing files during
configure time and if that fails, disabling aldor?
regards,
Franz
Bill Page
> ...
> Waldek, if you think it's wiser to ask the user to explicitly give the
> --enable-aldor switch at configure time, I don't oppose too much. I
> think the situation is different now.
>
I also think that it's wiser.
> Furthermore, William Stein allowed me to distribute a fricasaldor.spkg
> (which would basically just contain libaxiom.al). So when we generate a
> new fricas release, we could upload an appropriate .spkg and also ask
> users (or have a script) to fetch that spkg from
> http://sagemath.org/packages/optional/. They would thus get libaxiom.al
> from there and can (without much build time) complete the necessary
> compilation of just the runtime files locally.
>
Excellent.
> Ralf
>
> PS: Just give me a few more days to finish the infrastructure stuff.
>
+1 :-)
Thanks Ralf!
Regards,
Bill Page.
Ralf Hemmecke
>> and only build when explicitely requested.
> Just a thought from a non-autoconf-expert.
> Currently the aldor interface is enabled only if aldor is present.
> This indicates that the user is somehow interested in aldor
> and might want to have it available in fricas, too.
> Wouldn't it be possible to fetch the missing files during
> configure time and if that fails, disabling aldor?
That is a bit more difficult, because I'd have to create the
$build_dir/src/aldor directory before the corresponding configure part
creates it anyway by writing the Makefiles there.
But I believe that Waldek was against it (it's certainly somewhere in
the archive). I'm not currently interested in persueing this direction
again.
What would be much better... the necessary files, namely
https://svn.origo.ethz.ch/algebraist/trunk/aldor/lib/libax0/axiom.as
https://svn.origo.ethz.ch/algebraist/trunk/aldor/lib/libax0/axextend.as
https://svn.origo.ethz.ch/algebraist/trunk/aldor/lib/libax0/axlit.as
https://svn.origo.ethz.ch/algebraist/trunk/aldor/lib/libax0/stub.as
https://svn.origo.ethz.ch/algebraist/trunk/aldor/lib/libax0/minimach.as
https://svn.origo.ethz.ch/algebraist/trunk/aldor/lib/axllib/src/lang.as
and
$(ALDORROOT)/lib/libfoam.al(runtime.ao)
would be released under a modified BSD license. Until this happens
people should not complain to me but to people who have the strength to
really do something to remove this pain. Anything that requires
downloading files either at configure or build time is an imperfect
solution. :-(
Ralf