Re: [Reduce-algebra-developers] build problems on mac os x 10.6 (snow leopard)

[Arthur N. <ac...@ca...>]

On Sun, 8 Jan 2012, mark brethen wrote:
> An update to my previous post:
>
> Macports handles the configure and build phases; all that needs to be 
passed are the arguments, it handles all the environment variables. One 
drawback is that I can't pass both --with-csl and --with-psl. So I had to 
override macports' configure phase. At first, I used:
>
> configure {
>  system -W ${worksrcpath} "./configure --prefix=${prefix} --with-csl ; ./configure --with-psl"
> }
>

As regards this I might note that the CSL and PSL variants of Reduce are 
for almost all purposes ALTERNATIVES rather than being things where 
installing one should automatically install the other. I guess there are 
three sorts of people in the world:
(1) Developers who are not just keen on getting their Reduce package in 
good condition, but who wish to be properly careful that the code involved 
is compatible with both the CSL and the PSL-supported variants. Of all the 
people who download and install Reduce I believe these count as a tiny 
minority.
(2) General users. They only need EITHER the PSL or the CSL based version 
of Reduce. Installing both just clutters up their disc space, makes the 
installation process slower and in general causes confusion! There is of 
course a potential uncertainty as to which to choose, but even if theyu 
install both to start with after a while tidiness will cause them to 
de-install the one that does not suit them.
(3) Extreme users. Some (but ony a VERY FEW) users will be trying 
calculations so large that they fail on one or other version. Or they may 
trigger a bug in one or other version, or performance may differ enough to 
be critical. These people may want both versions installed so thay can 
switch between them as thay hit issues. Again I believe these are a tiny 
(but obviosuly important) minority.

So I might have expected that one Macport package for the CSL and one for 
the PSL version would be more use for customers as well as less strain for 
you.


> but I didn't realize that by overriding macports configure, the 
environment variables didn't get passed (as seen in macports' log):
>
> :info:configure in-place build attempt = yes
> :info:configure configure: host=x86_64-apple-darwin11.2.0 args= '--prefix=/opt/local' '--with-csl'
> :info:configure configure: Will build in the x86_64-mac_10.7_lion-darwin11.2.0 subdirectory
> :info:configure configure: +++ Will build in /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk/cslbuild/x86_64-mac_10.7_lion-darwin11.2.0
> :info:configure configure: passcc =
> :info:configure configure: doconfig = /bin/sh /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk/configure  '--prefix=/opt/local' '--with-csl'    CPPFLAGS="" CFLAGS=""             CXXFLAGS="" LDFLAGS=""              --with-build="x86_64-mac_10.7_lion-darwin11.2.0" --with-pslbuild="x86_64-mac_10.7_lion-darwin11.2.0"
> :info:configure configure: Absolute path to source directory = /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk
> :info:configure checking build system type... x86_64-apple-darwin11.2.0
> :info:configure checking host system type... x86_64-apple-darwin11.2.0
> :info:configure in-place build attempt = no
> :info:configure configure: host=x86_64-apple-darwin11.2.0 args= '--prefix=/opt/local' '--with-csl' 'CPPFLAGS=' 'CFLAGS=' 'CXXFLAGS=' 'LDFLAGS=' '--with-build=x86_64-mac_10.7_lion-darwin11.2.0' '--with-pslbuild=x86_64-mac_10.7_lion-darwin11.2.0'
> :info:configure configure: Will build in the x86_64-mac_10.7_lion-darwin11.2.0 subdirectory
> :info:configure configure: +++ standard build case, srcdir = /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk = /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk
>
> Amazingly, this compiled on lion, but did not for the fellow who had 
that script on his 10.6 machine. So I added the variables like so:
>
> configure {
>  system -W ${worksrcpath} \
>     "CC=${configure.cc} \
>      CFLAGS=\"${configure.cflags} [get_canonical_archflags cc]\" \
>      CXX=${configure.cxx} \
>      CXXFLAGS=\"${configure.cxxflags} [get_canonical_archflags cxx]\" \
>      CPPFLAGS=\"${configure.cppflags}\" \
>      LDFLAGS=\"${configure.ldflags} [get_canonical_archflags ld]\" \
>      ./configure --prefix=${prefix} --with-csl ; ./configure --with-psl"
> }
>
> Now the correct environment is set, as you can see:
>
> :info:configure configure: Absolute path to source directory = /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk
> :info:configure checking build system type... x86_64-apple-darwin11.2.0
> :info:configure checking host system type... x86_64-apple-darwin11.2.0
> :info:configure in-place build attempt = yes
> :info:configure configure: host=x86_64-apple-darwin11.2.0 args= '--prefix=/opt/local' '--with-csl'
> :info:configure configure: Will build in the x86_64-mac_10.7_lion-darwin11.2.0 subdirectory
> :info:configure configure: +++ Will build in /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk/cslbuild/x86_64-mac_10.7_lion-darwin11.2.0
> :info:configure configure: passcc = CC="/Developer/usr/bin/clang" CPP="" CXX="/Developer/usr/bin/clang++" CXXCPP=""
> :info:configure configure: doconfig = /bin/sh /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk/configure  '--prefix=/opt/local' '--with-csl'    CPPFLAGS="-I/opt/local/include" CFLAGS="-O2 -arch x86_64"             CXXFLAGS="-O2 -arch x86_64" LDFLAGS="-L/opt/local/lib -arch x86_64"             CC="/Developer/usr/bin/clang" CPP="" CXX="/Developer/usr/bin/clang++" CXXCPP="" --with-build="x86_64-mac_10.7_lion-darwin11.2.0" --with-pslbuild="x86_64-mac_10.7_lion-darwin11.2.0"
> :info:configure configure: Absolute path to source directory = /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk
> :info:configure checking build system type... x86_64-apple-darwin11.2.0
> :info:configure checking host system type... x86_64-apple-darwin11.2.0
> :info:configure in-place build attempt = no
> :info:configure configure: host=x86_64-apple-darwin11.2.0 args= '--prefix=/opt/local' '--with-csl' 'CPPFLAGS=-I/opt/local/include' 'CFLAGS=-O2 -arch x86_64' 'CXXFLAGS=-O2 -arch x86_64' 'LDFLAGS=-L/opt/local/lib -arch x86_64' 'CC=/Developer/usr/bin/clang' 'CPP=' 'CXX=/Developer/usr/bin/clang++' 'CXXCPP=' '--with-build=x86_64-mac_10.7_lion-darwin11.2.0' '--with-pslbuild=x86_64-mac_10.7_lion-darwin11.2.0'
> :info:configure configure: Will build in the x86_64-mac_10.7_lion-darwin11.2.0 subdirectory
> :info:configure configure: +++ standard build case, srcdir = /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk = /opt/local/var/macports/build/_Users_marbre_ports_math_reduce/reduce/work/trunk
>
> It builds on lion, but with the following errors:
>
> :info:configure configure: Building for Macintosh/Darwin with X
> :info:configure llvm-gcc-4.2: error trying to exec '/usr/bin/../llvm-gcc-4.2/bin/powerpc-apple-darwin11-llvm-gcc-4.2': execvp: No such file or directory
> :info:configure lipo: can't figure out the architecture type of: /var/tmp//ccAr5REE.out
> :info:configure llvm-gcc-4.2: error trying to exec '/usr/bin/../llvm-gcc-4.2/bin/powerpc-apple-darwin11-llvm-gcc-4.2': execvp: No such file or directory
> :info:configure conftest.c:1:19: error: stdio.h: No such file or directory
> :info:configure conftest.c: In function 'main':
> :info:configure conftest.c:3: warning: incompatible implicit declaration of built-in function 'printf'
> :info:configure lipo: can't figure out the architecture type of: /var/tmp//ccjtXcMt.out
> :info:configure llvm-gcc-4.2: error trying to exec '/usr/bin/../llvm-gcc-4.2/bin/powerpc-apple-darwin11-llvm-gcc-4.2': execvp: No such file or directory
> :info:configure conftest.c:1:19: error: stdio.h: No such file or directory
> :info:configure conftest.c: In function 'main':
> :info:configure conftest.c:3: warning: incompatible implicit declaration of built-in function 'printf'
> :info:configure lipo: can't figure out the architecture type of: /var/tmp//ccD75kI3.out
> :info:configure configure: fat binary creation will probably NOT be possible
>
>
> These errors were in the previous log as well. I don't know if this is 
because those instructions work for a 10.6 installation, but not 10.7? 
And there's still those explicit "gcc" calls, as I pointed out in my 
previous message.

Your extract showing "errors" is in fact not terribly helpful to me even 
if I wanted to check it! It does not provide context. Is this an extract 
from a config.log? You will be aware that configure does all sorts of 
things that cause "errors" in that when it is checking for the 
availability of something it tries using it, and if the attempt does or 
does not succeed it sets its various flags. So from what you send I can 
not tell if the messages you send are merely the normal process of 
configuration when it checks to see if llvm-gcc-4.2 is installed as one of 
the possible compilers of interest, or of it is an Error with a capital 
letter. In general when configure is run and something fails one looks in 
config.log and can see the exact command that was invoked and the contents 
of the file that was being presented for compilation. It is lengthy and 
messy and you need (as I say) to distinguish between compilations that 
fail merely incidentally because they are checking for some (optional) 
feature that is not present on the machine concerned, and errors that are 
indications of true disaster.



I have just checked in a new set of configure files. If the person who 
invoked configure does so with CC or CXX set (to values other than gcc and 
g++) then it skips the Mac code that uses gcc explicitly and tries to
identity which SDK is available. The reasoning there is that if anybody is
going to force selection of some particular C compiler then it becomes 
THEIR responsibility to force in values of CFLAGS, LDFLAGS, LIBS and 
anything else that could matter, including special flags to select an SDK 
or a word length.

That I think purs responsibility more clearly where it has to lie! I hope 
I got my changes correct - let me know if I messed up!!!


>
> -Mark

Arthur