hi there, I just tried to compile boost.python v2 on irix using the 'mipspro' toolset. The compilation failed on the file 'pickle_support.o' with the error(s) cc-3354 CC: ERROR File = /homes/Developer/sseefel/boost/boost/mpl/aux_/preprocessed/plain/arg.hpp, Line = 18 A non-integral operation is not allowed in a nontype template argument. BOOST_STATIC_ASSERT(!is_void_<type>::value); ^ cc-1070 CC: ERROR File = /homes/Developer/sseefel/boost/boost/mpl/aux_/preprocessed/plain/arg.hpp, Line = 18 The indicated type is incomplete. BOOST_STATIC_ASSERT(!is_void_<type>::value); Beside that, I find the build system not very transparent, so tracking things down is non-obvious. The use of jam isn't particularly helpful. For example I just realized from the error logs that it is using the "-mips3" compiler flag, while I'd like "-mips4". I couldn't find either docs nor code that made clear how to change that flag. Any help would be highly appreciated ! Regards, Stefan
[2002-10-07] Stefan Seefeld wrote:
hi there,
Beside that, I find the build system not very transparent, so tracking things down is non-obvious. The use of jam isn't particularly helpful. For example I just realized from the error logs that it is using the "-mips3" compiler flag, while I'd like "-mips4". I couldn't find either docs nor code that made clear how to change that flag.
Any help would be highly appreciated !
bjam ... "-sBUILD=<architecture>mips4" If you look at the *-tools.jam files the build system looks fairly transparent. All the commands and mapping for flags are right there. But if you look at other *.jam files it certainly is not transparent, and it wasn't meant to be. -- grafik - Don't Assume Anything -- rrivera@acm.org - grafik@redshift-software.com -- 102708583@icq - Grafik666@AIM - Grafik@jabber.org
Rene Rivera wrote:
bjam ... "-sBUILD=<architecture>mips4"
thanks !
If you look at the *-tools.jam files the build system looks fairly transparent.
you are assuming that people know jam syntax, aren't you ?
All the commands and mapping for flags are right there. But if you look at other *.jam files it certainly is not transparent, and it wasn't meant to be.
Well, with transparency I meant (among others) that it should be easy to see that the current build uses '-mips3'. Beside, most people are trained with 'make', so every deviation should be well documented somewhere (jam files vs. Makefiles). But may be I just didn't look hard enough... In my own ('make' based) build systems I make sure all commands are dumped to stdout (as is the default anyways) when I invoke 'make', and something like in your build system, i.e. messages of the form 'compiling foobar.cc' if I invoke 'make -s', where I explicitely ask the build to be less verbose. Best regards, Stefan
[2002-10-07] Stefan Seefeld wrote:
Rene Rivera wrote:
bjam ... "-sBUILD=<architecture>mips4"
thanks !
If you look at the *-tools.jam files the build system looks fairly transparent.
you are assuming that people know jam syntax, aren't you ?
No. Just that they know how to use the compiler they are intrested in. Because most things in Boost.Build are fairly wordy it's fairly easy to see which flags are used depending on the circumstances. Like the -mips4 vs. -mips3 flags. In the toolset file you have: flags mipspro CFLAGS <architecture>mips3 : -mips3 ; flags mipspro CFLAGS <architecture>mips4 : -mips4 ; That seems very clear where the flags come from. Now yes I must say that knowing that to specify that you use the "-sBUILD=..." is another problem, which requires reading way more build system docs. We know the doc problem, I pointed it the first time I got involved with the build system. But we don't have the time to be complete :-(
All the commands and mapping for flags are right there. But if you look at other *.jam files it certainly is not transparent, and it wasn't meant to be.
Well, with transparency I meant (among others) that it should be easy to see that the current build uses '-mips3'. Beside, most people are trained with 'make', so every deviation should be well documented somewhere (jam files vs. Makefiles). But may be I just didn't look hard enough...
flags mipspro CFLAGS <architecture>native : -mips3 ;
In my own ('make' based) build systems I make sure all commands are dumped to stdout (as is the default anyways) when I invoke 'make', and something like in your build system, i.e. messages of the form 'compiling foobar.cc' if I invoke 'make -s', where I explicitely ask the build to be less verbose.
Jam and Boost.Build take the opposite approach. The reason is simple... after you know how to use the system you don't care about what commands it executes anymore. So we default to less verbosity at the outset. If you want to see the commands it's going to execute try the "-n" flag. -- grafik - Don't Assume Anything -- rrivera@acm.org - grafik@redshift-software.com -- 102708583@icq - Grafik666@AIM - Grafik@jabber.org
Stefan Seefeld <stefan.seefeld@orthosoft.ca> writes:
hi there,
I just tried to compile boost.python v2 on irix using the 'mipspro' toolset. The compilation failed on the file 'pickle_support.o' with the error(s)
cc-3354 CC: ERROR File = /homes/Developer/sseefel/boost/boost/mpl/aux_/preprocessed/plain/arg.hpp, Line = 18 A non-integral operation is not allowed in a nontype template argument.
BOOST_STATIC_ASSERT(!is_void_<type>::value); ^
cc-1070 CC: ERROR File = /homes/Developer/sseefel/boost/boost/mpl/aux_/preprocessed/plain/arg.hpp, Line = 18 The indicated type is incomplete.
BOOST_STATIC_ASSERT(!is_void_<type>::value);
This was working on that compiler a few days ago. I'll see about a workaround.
Beside that, I find the build system not very transparent, so tracking things down is non-obvious. The use of jam isn't particularly helpful. For example I just realized from the error logs that it is using the "-mips3" compiler flag, while I'd like "-mips4". I couldn't find either docs nor code that made clear how to change that flag.
Yeah, we know there are some doc limitations we need to address. It's a tension between investing resources in the current system and developing the next one. We'll try to help out as best we can in the meantime. -- David Abrahams * Boost Consulting dave@boost-consulting.com * http://www.boost-consulting.com
--- Stefan Seefeld <stefan.seefeld@orthosoft.ca> wrote:
I just tried to compile boost.python v2 on irix using the 'mipspro' toolset. The compilation failed on the file 'pickle_support.o' with the error(s)
cc-3354 CC: ERROR File = /homes/Developer/sseefel/boost/boost/mpl/aux_/preprocessed/plain/arg.hpp, Line = 18 A non-integral operation is not allowed in a nontype template argument.
BOOST_STATIC_ASSERT(!is_void_<type>::value); ^
We are very actively maintaining the MIPSpro builds. As of 00:00PDT the build with a fresh CVS was successful (http://cci.lbl.gov/boost/). When did you check out your copy of the code? We are using MIPSpro versions 7.3.1.2m and 7.3.1.3m. What are you using? Ralf __________________________________________________ Do you Yahoo!? Faith Hill - Exclusive Performances, Videos & More http://faith.yahoo.com
The 8am build shows that the current CVS does indeed not work with IRIX. Until we have a work-around you can use the cvs -D option to check out the cvs from 0am PDT (see man cvs). Ralf --- "Ralf W. Grosse-Kunstleve" <rwgk@yahoo.com> wrote:
--- Stefan Seefeld <stefan.seefeld@orthosoft.ca> wrote:
I just tried to compile boost.python v2 on irix using the 'mipspro' toolset. The compilation failed on the file 'pickle_support.o' with the error(s)
__________________________________________________ Do you Yahoo!? Faith Hill - Exclusive Performances, Videos & More http://faith.yahoo.com
Ralf W. Grosse-Kunstleve wrote:
We are very actively maintaining the MIPSpro builds. As of 00:00PDT the build with a fresh CVS was successful (http://cci.lbl.gov/boost/). When did you check out your copy of the code?
just before compiling (and mailing to you) this morning.
We are using MIPSpro versions 7.3.1.2m and 7.3.1.3m. What are you using?
MIPSpro 7.3.1.3m too, here. And here are some more errors: I removed the pickling module from the Jam file (hopefully it isn't needed by any other module), and it stops at : Chmod1 libs/python/build/bin/libboost_python.so/mipspro/release/libboost_python.so FileClone libs/python/build/bin-stage/libboost_python.so Illegal option -- d Usage: cp [-aDfirRp] [-b size] [-tP -[e size]] file1 file2 cp [-aDfirRp] [-b size] [-tP -[e size]] file1 ... file2 dir cp -fdp libs/python/build/bin/libboost_python.so/mipspro/release/libboost_python.so libs/python/build/bin-stage/libboost_python.so Then I started on my own code, adding a Makefile...It stopped because it was looking for <cstdlib> et al. I compiled the examples using bjam, and was lucky to get some errors, since that permitted me to see what flags you use, i.e. I discovered by chance that you provide C++ compatibility headers. My transparency argument again... Regards, Stefan
Stefan Seefeld <seefeld@sympatico.ca> writes:
I removed the pickling module from the Jam file (hopefully it isn't needed by any other module), and it stops at :
Chmod1 libs/python/build/bin/libboost_python.so/mipspro/release/libboost_python.so FileClone libs/python/build/bin-stage/libboost_python.so Illegal option -- d Usage: cp [-aDfirRp] [-b size] [-tP -[e size]] file1 file2 cp [-aDfirRp] [-b size] [-tP -[e size]] file1 ... file2 dir
cp -fdp libs/python/build/bin/libboost_python.so/mipspro/release/libboost_python.so libs/python/build/bin-stage/libboost_python.so
Hmm, this part is due to Rene's new tagging stuff. Rene? In the meantime, you should have no problems with the RC_1_29_0 branch of the boost CVS using Boost.Build. -- David Abrahams * Boost Consulting dave@boost-consulting.com * http://www.boost-consulting.com
Stefan Seefeld <seefeld@sympatico.ca> writes:
I removed the pickling module from the Jam file (hopefully it isn't needed by any other module), and it stops at :
Chmod1
FileClone libs/python/build/bin-stage/libboost_python.so Illegal option -- d Usage: cp [-aDfirRp] [-b size] [-tP -[e size]] file1 file2 cp [-aDfirRp] [-b size] [-tP -[e size]] file1 ... file2 dir
cp -fdp
[2002-10-07] David Abrahams wrote: libs/python/build/bin/libboost_python.so/mipspro/release/libboost_python.so libs/python/build/bin/libboost_python.so/mipspro/release/libboost_python.so
libs/python/build/bin-stage/libboost_python.so
Hmm, this part is due to Rene's new tagging stuff. Rene?
It definately is... but not really new, just the use of FileClone by the stage rule.
In the meantime, you should have no problems with the RC_1_29_0 branch of the boost CVS using Boost.Build.
Beg to differ ;-) -- Those changes are also in the RC branch. Question that I need to answer is: What are the flags to "cp" so that it also copies symbolic links as is? -- grafik - Don't Assume Anything -- rrivera@acm.org - grafik@redshift-software.com -- 102708583@icq - Grafik666@AIM - Grafik@jabber.org
Rene Rivera <grafik666@redshift-software.com> writes:
libs/python/build/bin/libboost_python.so/mipspro/release/libboost_python.so
libs/python/build/bin-stage/libboost_python.so
Hmm, this part is due to Rene's new tagging stuff. Rene?
It definately is... but not really new, just the use of FileClone by the stage rule.
In the meantime, you should have no problems with the RC_1_29_0 branch of the boost CVS using Boost.Build.
Beg to differ ;-) -- Those changes are also in the RC branch.
?? I thought you were going to wait 'till I had tried this stuff before putting it in the release candidate. Or am I thinking of some other feature? -- David Abrahams * Boost Consulting dave@boost-consulting.com * http://www.boost-consulting.com
Rene Rivera <grafik666@redshift-software.com> writes:
[2002-10-07] David Abrahams wrote: libs/python/build/bin/libboost_python.so/mipspro/release/libboost_python.so
libs/python/build/bin-stage/libboost_python.so
Hmm, this part is due to Rene's new tagging stuff. Rene?
It definately is... but not really new, just the use of FileClone by the stage rule.
In the meantime, you should have no problems with the RC_1_29_0 branch of the boost CVS using Boost.Build.
Beg to differ ;-) -- Those changes are also in the RC branch.
?? I thought you were going to wait 'till I had tried this stuff before putting it in the release candidate. Or am I thinking of some other feature?
You are thinking of some other feature... The FileClone is what stage uses to copy the files. So if you use stage you'll see that problem. It's been that way for a while. It's the change that I did put into the RC before the main, because of the release instructions. -- grafik - Don't Assume Anything -- rrivera@acm.org - grafik@redshift-software.com -- 102708583@icq - Grafik666@AIM - Grafik@jabber.org
Rene Rivera <grafik666@redshift-software.com> writes:
[2002-10-07] David Abrahams wrote:
Rene Rivera <grafik666@redshift-software.com> writes:
libs/python/build/bin/libboost_python.so/mipspro/release/libboost_python.so
libs/python/build/bin-stage/libboost_python.so
Hmm, this part is due to Rene's new tagging stuff. Rene?
It definately is... but not really new, just the use of FileClone by the stage rule.
In the meantime, you should have no problems with the RC_1_29_0 branch of the boost CVS using Boost.Build.
Beg to differ ;-) -- Those changes are also in the RC branch.
?? I thought you were going to wait 'till I had tried this stuff before putting it in the release candidate. Or am I thinking of some other feature?
You are thinking of some other feature... The FileClone is what stage uses to copy the files. So if you use stage you'll see that problem. It's been that way for a while. It's the change that I did put into the RC before the main, because of the release instructions.
Ah, OK. Well I attached the IRIX man page for cp here: -- David Abrahams * Boost Consulting dave@boost-consulting.com * http://www.boost-consulting.com
Stefan Seefeld <stefan.seefeld@orthosoft.ca> writes:
hi there,
I just tried to compile boost.python v2 on irix using the 'mipspro' toolset. The compilation failed on the file 'pickle_support.o' with the error(s)
cc-3354 CC: ERROR File = /homes/Developer/sseefel/boost/boost/mpl/aux_/preprocessed/plain/arg.hpp, Line = 18 A non-integral operation is not allowed in a nontype template argument.
BOOST_STATIC_ASSERT(!is_void_<type>::value); ^
cc-1070 CC: ERROR File = /homes/Developer/sseefel/boost/boost/mpl/aux_/preprocessed/plain/arg.hpp, Line = 18 The indicated type is incomplete.
BOOST_STATIC_ASSERT(!is_void_<type>::value);
OK, this should be working now. Please give it a shot. -- David Abrahams * Boost Consulting dave@boost-consulting.com * http://www.boost-consulting.com
participants (5)
-
David Abrahams -
Ralf W. Grosse-Kunstleve -
Rene Rivera -
Stefan Seefeld -
Stefan Seefeld