From: Akos B. <edh...@cs...> - 2002-10-29 17:21:40
|
Hi guys, I wanted to check if something is a gcc issue in my code, so I wanted to recompile OpenSG with icc 6.0 (the intel compiler). I've installed icc, it seems to work. I run './configure --with-compiler=icc' and it seems to be OK more or less (for some reason it thinks it's a cross-compiler which is not true of course, but I don't think that's a major problem). When I try to compile the library itself I get tons of errors related to the standard gcc includes: (the dependency build goes fine, and then:) /local/work/OpenSG-CVS/OpenSG/Common/common.i686-pc-linux-gnu-icc.mk:125: Got Compiler Version : 600 make[1]: Entering directory `/local/work/OpenSG-CVS/OpenSG/Builds/i686-pc-linux-gnu-icc' /local/work/OpenSG-CVS/OpenSG/Common/common.i686-pc-linux-gnu-icc.mk:125: Got Compiler Version : 600 make[2]: Entering directory `/local/work/OpenSG-CVS/OpenSG/Builds/i686-pc-linux-gnu-icc/BaseLib' /local/work/OpenSG-CVS/OpenSG/Common/commonBuildDetect.mk:43: Detected DBG /local/work/OpenSG-CVS/OpenSG/Common/commonBuildDetect.mk:59: Detected Shared Lib /local/work/OpenSG-CVS/OpenSG/Common/common.i686-pc-linux-gnu-icc.mk:125: Got Compiler Version : 600 using BD = /local/work/OpenSG-CVS/OpenSG/Source icc -w2 -wd981,424,76 -D_GNU_SOURCE -DQT_CLEAN_NAMESPACE -DOSG_WITH_GLUT -DOSG_WITH_QT -DOSG_WITH_JPG -DOSG_WITH_PNG -DOSG_WITH_MNG -DOSG_SUPPORT_NO_GEO_INTERFACE=1 -D_OSG_HAVE_CONFIGURED_H_ -D__INTEL_COMPILER_VERSION=600 -DQT_NO_XINERAMA -DQT_NO_XRENDER -DQT_NO_XFTFREETYPE -DQT_NO_XKB -DQT_NO_SM_SUPPORT -DQT_NO_IMAGEIO_MNG -DQT_NO_IMAGEIO_JPEG -DQT_NO_STYLE_AQUA -DQT_NO_STYLE_MAC -DQT_NO_STYLE_INTERLACE -DQT_NO_STYLE_COMPACT -Tnoauto -KPIC -Kc++eh -Krtti -ansi -g -DOSG_DEBUG -inline_debug_info -wd981,424,76 -c -I"../Base" -I$BD/Base/Base -I$BD/Base/Field -I$BD/Base/Functors -I$BD/Base/Network/Base -I$BD/Base/Network/Socket -I$BD/Base/StringConversion \ -Iobj-dbg -I. \ -o obj-dbg/OSGBarrier.o /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGBarrier.cpp /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGBarrier.cpp /usr/include/stdlib.h(93): error #77: this declaration has no storage class or type specifier __BEGIN_NAMESPACE_STD ^ /usr/include/stdlib.h(95): error: expected a ";" typedef struct ^ /usr/include/stdlib.h(110): error: identifier "__END_NAMESPACE_STD" is undefined __END_NAMESPACE_STD ^ /usr/include/stdlib.h(115): error: expected a ";" __extension__ typedef struct ^ /usr/include/stdlib.h(121): error #77: this declaration has no storage class or type specifier __END_NAMESPACE_C99 ^ [...] and hundrends of more errors. I made a simple testprogram, which just includes stdlib.h and even that doesn't compile, showing similar errors. Is this a general incompatibility issue between gcc header files and icc? I know there's some incompatibility, but one should be able to mix C source files, and compile C++ libraries/projects from scratch via icc. I've checked the /opt/intel dirs, and they do ship a number of include files, but those don't contain either stdlib.h or stdio.h so I would assume these are OK to use with icc... :-\ And I _think_ icc on Linux is a supported platform for OpenSG, isn't that so? I'm probably overlooking something trivial, but I can't seem to find it... Platform: RedHat 8.0, gcc 3.2, icc: $ rpm -qa |grep intel intel-isubh6-6.0-139 intel-icc6-6.0-139 intel-ildb6-6.0-189 Any ideas/suggestions? Akos Balazs |
From: Gerrit V. <vo...@ca...> - 2002-10-30 01:36:23
|
Hi, up to redhat 7.3 it works, if 8.0 introduces some problems I have to check, but I have to install it so it might take a while. gerrit On Wed, 2002-10-30 at 01:21, Akos Balazs wrote: > Hi guys, > > I wanted to check if something is a gcc issue in my code, so I wanted to > recompile OpenSG with icc 6.0 (the intel compiler). > > I've installed icc, it seems to work. > > I run './configure --with-compiler=icc' and it seems to be OK more or less > (for some reason it thinks it's a cross-compiler which is not true of > course, but I don't think that's a major problem). > > When I try to compile the library itself I get tons of errors related to > the standard gcc includes: > > (the dependency build goes fine, and then:) > > /local/work/OpenSG-CVS/OpenSG/Common/common.i686-pc-linux-gnu-icc.mk:125: > Got Compiler Version : 600 > make[1]: Entering directory > `/local/work/OpenSG-CVS/OpenSG/Builds/i686-pc-linux-gnu-icc' > /local/work/OpenSG-CVS/OpenSG/Common/common.i686-pc-linux-gnu-icc.mk:125: > Got Compiler Version : 600 > make[2]: Entering directory > `/local/work/OpenSG-CVS/OpenSG/Builds/i686-pc-linux-gnu-icc/BaseLib' > /local/work/OpenSG-CVS/OpenSG/Common/commonBuildDetect.mk:43: Detected DBG > /local/work/OpenSG-CVS/OpenSG/Common/commonBuildDetect.mk:59: Detected > Shared Lib > /local/work/OpenSG-CVS/OpenSG/Common/common.i686-pc-linux-gnu-icc.mk:125: > Got Compiler Version : 600 > using BD = /local/work/OpenSG-CVS/OpenSG/Source > icc -w2 -wd981,424,76 -D_GNU_SOURCE -DQT_CLEAN_NAMESPACE > -DOSG_WITH_GLUT -DOSG_WITH_QT -DOSG_WITH_JPG -DOSG_WITH_PNG -DOSG_WITH_MNG > -DOSG_SUPPORT_NO_GEO_INTERFACE=1 -D_OSG_HAVE_CONFIGURED_H_ > -D__INTEL_COMPILER_VERSION=600 -DQT_NO_XINERAMA -DQT_NO_XRENDER > -DQT_NO_XFTFREETYPE -DQT_NO_XKB -DQT_NO_SM_SUPPORT -DQT_NO_IMAGEIO_MNG > -DQT_NO_IMAGEIO_JPEG -DQT_NO_STYLE_AQUA -DQT_NO_STYLE_MAC > -DQT_NO_STYLE_INTERLACE -DQT_NO_STYLE_COMPACT -Tnoauto -KPIC -Kc++eh > -Krtti -ansi > -g -DOSG_DEBUG -inline_debug_info -wd981,424,76 -c -I"../Base" > -I$BD/Base/Base -I$BD/Base/Field -I$BD/Base/Functors > -I$BD/Base/Network/Base -I$BD/Base/Network/Socket > -I$BD/Base/StringConversion \ > -Iobj-dbg -I. \ > -o obj-dbg/OSGBarrier.o > /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGBarrier.cpp > /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGBarrier.cpp > /usr/include/stdlib.h(93): error #77: this declaration has no storage > class or type specifier > __BEGIN_NAMESPACE_STD > ^ > > /usr/include/stdlib.h(95): error: expected a ";" > typedef struct > ^ > > /usr/include/stdlib.h(110): error: identifier "__END_NAMESPACE_STD" is > undefined __END_NAMESPACE_STD > ^ > > /usr/include/stdlib.h(115): error: expected a ";" > __extension__ typedef struct > ^ > > /usr/include/stdlib.h(121): error #77: this declaration has no storage > class or > type specifier > __END_NAMESPACE_C99 > ^ > > > [...] > > and hundrends of more errors. > > I made a simple testprogram, which just includes stdlib.h and even that > doesn't compile, showing similar errors. > > Is this a general incompatibility issue between gcc header files and icc? > > I know there's some incompatibility, but one should be able to mix C > source files, and compile C++ libraries/projects from scratch via icc. > > I've checked the /opt/intel dirs, and they do ship a number of include > files, but those don't contain either stdlib.h or stdio.h so I would > assume these are OK to use with icc... :-\ > > And I _think_ icc on Linux is a supported platform for OpenSG, isn't that > so? > > > I'm probably overlooking something trivial, but I can't seem to find it... > > Platform: RedHat 8.0, gcc 3.2, icc: > > $ rpm -qa |grep intel > intel-isubh6-6.0-139 > intel-icc6-6.0-139 > intel-ildb6-6.0-189 > > > Any ideas/suggestions? > > Akos Balazs > > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > Opensg-users mailing list > Ope...@li... > https://lists.sourceforge.net/lists/listinfo/opensg-users |
From: Gerrit V. <vo...@ca...> - 2002-10-30 08:40:54
|
Hi, unfortunately reproducable, I'll see what can be done sorry for the bad news gerrit On Wed, 2002-10-30 at 09:36, Gerrit Voss wrote: > > Hi, > > > up to redhat 7.3 it works, if 8.0 introduces some problems > I have to check, but I have to install it so it might take > a while. > > gerrit > > On Wed, 2002-10-30 at 01:21, Akos Balazs wrote: > > Hi guys, > > > > I wanted to check if something is a gcc issue in my code, so I wanted to > > recompile OpenSG with icc 6.0 (the intel compiler). > > > > I've installed icc, it seems to work. > > > > I run './configure --with-compiler=icc' and it seems to be OK more or less > > (for some reason it thinks it's a cross-compiler which is not true of > > course, but I don't think that's a major problem). > > > > When I try to compile the library itself I get tons of errors related to > > the standard gcc includes: > > > > (the dependency build goes fine, and then:) > > > > /local/work/OpenSG-CVS/OpenSG/Common/common.i686-pc-linux-gnu-icc.mk:125: > > Got Compiler Version : 600 > > make[1]: Entering directory > > `/local/work/OpenSG-CVS/OpenSG/Builds/i686-pc-linux-gnu-icc' > > /local/work/OpenSG-CVS/OpenSG/Common/common.i686-pc-linux-gnu-icc.mk:125: > > Got Compiler Version : 600 > > make[2]: Entering directory > > `/local/work/OpenSG-CVS/OpenSG/Builds/i686-pc-linux-gnu-icc/BaseLib' > > /local/work/OpenSG-CVS/OpenSG/Common/commonBuildDetect.mk:43: Detected DBG > > /local/work/OpenSG-CVS/OpenSG/Common/commonBuildDetect.mk:59: Detected > > Shared Lib > > /local/work/OpenSG-CVS/OpenSG/Common/common.i686-pc-linux-gnu-icc.mk:125: > > Got Compiler Version : 600 > > using BD = /local/work/OpenSG-CVS/OpenSG/Source > > icc -w2 -wd981,424,76 -D_GNU_SOURCE -DQT_CLEAN_NAMESPACE > > -DOSG_WITH_GLUT -DOSG_WITH_QT -DOSG_WITH_JPG -DOSG_WITH_PNG -DOSG_WITH_MNG > > -DOSG_SUPPORT_NO_GEO_INTERFACE=1 -D_OSG_HAVE_CONFIGURED_H_ > > -D__INTEL_COMPILER_VERSION=600 -DQT_NO_XINERAMA -DQT_NO_XRENDER > > -DQT_NO_XFTFREETYPE -DQT_NO_XKB -DQT_NO_SM_SUPPORT -DQT_NO_IMAGEIO_MNG > > -DQT_NO_IMAGEIO_JPEG -DQT_NO_STYLE_AQUA -DQT_NO_STYLE_MAC > > -DQT_NO_STYLE_INTERLACE -DQT_NO_STYLE_COMPACT -Tnoauto -KPIC -Kc++eh > > -Krtti -ansi > > -g -DOSG_DEBUG -inline_debug_info -wd981,424,76 -c -I"../Base" > > -I$BD/Base/Base -I$BD/Base/Field -I$BD/Base/Functors > > -I$BD/Base/Network/Base -I$BD/Base/Network/Socket > > -I$BD/Base/StringConversion \ > > -Iobj-dbg -I. \ > > -o obj-dbg/OSGBarrier.o > > /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGBarrier.cpp > > /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGBarrier.cpp > > /usr/include/stdlib.h(93): error #77: this declaration has no storage > > class or type specifier > > __BEGIN_NAMESPACE_STD > > ^ > > > > /usr/include/stdlib.h(95): error: expected a ";" > > typedef struct > > ^ > > > > /usr/include/stdlib.h(110): error: identifier "__END_NAMESPACE_STD" is > > undefined __END_NAMESPACE_STD > > ^ > > > > /usr/include/stdlib.h(115): error: expected a ";" > > __extension__ typedef struct > > ^ > > > > /usr/include/stdlib.h(121): error #77: this declaration has no storage > > class or > > type specifier > > __END_NAMESPACE_C99 > > ^ > > > > > > [...] > > > > and hundrends of more errors. > > > > I made a simple testprogram, which just includes stdlib.h and even that > > doesn't compile, showing similar errors. > > > > Is this a general incompatibility issue between gcc header files and icc? > > > > I know there's some incompatibility, but one should be able to mix C > > source files, and compile C++ libraries/projects from scratch via icc. > > > > I've checked the /opt/intel dirs, and they do ship a number of include > > files, but those don't contain either stdlib.h or stdio.h so I would > > assume these are OK to use with icc... :-\ > > > > And I _think_ icc on Linux is a supported platform for OpenSG, isn't that > > so? > > > > > > I'm probably overlooking something trivial, but I can't seem to find it... > > > > Platform: RedHat 8.0, gcc 3.2, icc: > > > > $ rpm -qa |grep intel > > intel-isubh6-6.0-139 > > intel-icc6-6.0-139 > > intel-ildb6-6.0-189 > > > > > > Any ideas/suggestions? > > > > Akos Balazs > > > > > > > > ------------------------------------------------------- > > This sf.net email is sponsored by:ThinkGeek > > Welcome to geek heaven. > > http://thinkgeek.com/sf > > _______________________________________________ > > Opensg-users mailing list > > Ope...@li... > > https://lists.sourceforge.net/lists/listinfo/opensg-users > > > > > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > Opensg-users mailing list > Ope...@li... > https://lists.sourceforge.net/lists/listinfo/opensg-users |
From: Akos B. <edh...@cs...> - 2002-10-30 09:47:09
|
Hi Gerrit, On 30 Oct 2002, Gerrit Voss wrote: > > Hi, > > unfortunately reproducable, I'll see what can be done > > sorry for the bad news > gerrit It's bad news in the sense that it doesn't work, but good news (for me :) in the sense that it's not me who f***ed up something this time. ;) As I wrote earlier I actually suspect it to be more of an icc 6.0 vs gcc 3.2 headers issue, try writing a very small testprogram which includes stdio.h or stdlib.h and it will fail aswell. ;( I don't know, why it thinks it's cross-compiling though. I can't log in to intel's support pages because their web-system kinda forgot to send me some license and other files (I've already bugged them about it but no reply so far :( ) but as soon as I can I'll try to ask them about it if it's not resolved by that time... Thanks for your efforts, Akos Balazs |
From: Gerrit V. <vo...@ca...> - 2002-10-30 10:40:03
|
Hi, On Wed, 2002-10-30 at 17:44, Akos Balazs wrote: > Hi Gerrit, > > > On 30 Oct 2002, Gerrit Voss wrote: > > > > > Hi, > > > > unfortunately reproducable, I'll see what can be done > > > > sorry for the bad news > > gerrit > > It's bad news in the sense that it doesn't work, but good news (for me :) > in the sense that it's not me who f***ed up something this time. ;) not as bad as the one to come ;-), I'll refixed all the substitute_header, and just let the compiler run, so far Base is done, looks ok. I'll post it soon because you probably have to compile something else ;-) Ok, now the next tricky point : If you try ldd /usr/lib/libGLU.so and get an output which includes libstdc++.so.5 you must rebuild libGLU.so otherwise you catch two c++ stdlibs at runtime and this is no fun at all ;-) Probably I'll send intel my patchset too ;-). As soon as the compilation is finished I'll post the headers. gerrit > > As I wrote earlier I actually suspect it to be more of an icc 6.0 vs gcc > 3.2 headers issue, try writing a very small testprogram which includes > stdio.h or stdlib.h and it will fail aswell. ;( > > I don't know, why it thinks it's cross-compiling though. > > I can't log in to intel's support pages because their web-system kinda > forgot to send me some license and other files (I've already bugged them > about it but no reply so far :( ) but as soon as I can I'll try to ask > them about it if it's not resolved by that time... > > > Thanks for your efforts, > Akos Balazs > > > > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > Opensg-users mailing list > Ope...@li... > https://lists.sourceforge.net/lists/listinfo/opensg-users |
From: Akos B. <edh...@cs...> - 2002-10-30 11:24:46
|
Hi Gerrit, On 30 Oct 2002, Gerrit Voss wrote: > > > unfortunately reproducable, I'll see what can be done > > > > > > sorry for the bad news > > > gerrit > > > > It's bad news in the sense that it doesn't work, but good news (for me :) > > in the sense that it's not me who f***ed up something this time. ;) > > not as bad as the one to come ;-), I'll refixed all the > substitute_header, and just let the compiler run, so far > Base is done, looks ok. I'll post it soon because you probably > have to compile something else ;-) What's this substitute_header stuff? What I don't understand is that if you have a minimalistic program, say: #include <stdlib.h> int main(int argc, char **argv) { std::cerr << " hello " << std::endl; return 0; } you'll get lotsof errors because of stdlib.h during compilation with icc. That's why I thought it to be a gcc 3.2 vs icc 6.0 issue... > Ok, now the next tricky point : > > If you try ldd /usr/lib/libGLU.so and get an output which > includes libstdc++.so.5 you must rebuild libGLU.so otherwise > you catch two c++ stdlibs at runtime and this is no fun at all ;-) Yeah, libstdc++.so.5 is there, alright... :( seems I'll have to rebuild this one too. Sometimes I _do_ love Linux. :-) > Probably I'll send intel my patchset too ;-). > > As soon as the compilation is finished I'll post the headers. What patchset are you talking about? Do you mean you actually have to patch the system-wide headers? Ouch... :\ Regards, Akos Balazs |
From: Gerrit V. <vo...@ca...> - 2002-10-30 11:54:57
Attachments:
subst_headers_rh_8.0.tgz
hash_map.diff
|
Hi, On Wed, 2002-10-30 at 19:24, Akos Balazs wrote: > Hi Gerrit, > > > On 30 Oct 2002, Gerrit Voss wrote: > > > > > unfortunately reproducable, I'll see what can be done > > > > > > > > sorry for the bad news > > > > gerrit > > > > > > It's bad news in the sense that it doesn't work, but good news (for me :) > > > in the sense that it's not me who f***ed up something this time. ;) > > > > not as bad as the one to come ;-), I'll refixed all the > > substitute_header, and just let the compiler run, so far > > Base is done, looks ok. I'll post it soon because you probably > > have to compile something else ;-) > > What's this substitute_header stuff? What I don't understand is that if > you have a minimalistic program, say: > > #include <stdlib.h> > > int main(int argc, char **argv) > { > std::cerr << " hello " << std::endl; > return 0; > } > > you'll get lotsof errors because of stdlib.h during compilation with icc. > That's why I thought it to be a gcc 3.2 vs icc 6.0 issue... that's because intel replaces the cdefs.h file which includes all the new namespace defines from gcc 3.2. That's I guess 90% of the error you see. > > > Ok, now the next tricky point : > > > > If you try ldd /usr/lib/libGLU.so and get an output which > > includes libstdc++.so.5 you must rebuild libGLU.so otherwise > > you catch two c++ stdlibs at runtime and this is no fun at all ;-) > > Yeah, libstdc++.so.5 is there, alright... :( seems I'll have to > rebuild this one too. Sometimes I _do_ love Linux. :-) > > > Probably I'll send intel my patchset too ;-). > > > > As soon as the compilation is finished I'll post the headers. > > What patchset are you talking about? Do you mean you actually have to > patch the system-wide headers? Ouch... :\ > attached you find one tgz archive which contains a new substitute_header directory, just move the old one away and untar the archive and one patch for include/hash_map. Both only affects the intel/ia32/ installation, so no need to change the system. hope it works, at least it compiles for me, now it's time to rebuild libGLU.so ;-) have fun, gerrit |
From: Akos B. <edh...@cs...> - 2002-10-30 12:42:23
|
Hi Gerrit, > > What's this substitute_header stuff? What I don't understand is that if > > you have a minimalistic program, say: [...] > > > > you'll get lotsof errors because of stdlib.h during compilation with icc. > > That's why I thought it to be a gcc 3.2 vs icc 6.0 issue... > > that's because intel replaces the cdefs.h file which includes > all the new namespace defines from gcc 3.2. That's I guess 90% of > the error you see. Oh, I get it now. I was still half asleep I guess... :) > > > Probably I'll send intel my patchset too ;-). > > > > > > As soon as the compilation is finished I'll post the headers. > > > > What patchset are you talking about? Do you mean you actually have to > > patch the system-wide headers? Ouch... :\ > > > > attached you find one tgz archive which contains a new substitute_header > directory, just move the old one away and untar the archive and one > patch for include/hash_map. Both only affects the intel/ia32/ > installation, so no need to change the system. Thanks a lot, now it seems to compile. > hope it works, at least it compiles for me, now it's time to rebuild > libGLU.so ;-) Could you send this to me aswell if you're building on RedHat 8.0 with gcc? (I don't want my system-wide stuff to be compiled with icc, although in the case of C sources it probably doesn't (and shouldn't) matter...) Thanks again, Akos |
From: Gerrit V. <vo...@ca...> - 2002-10-31 02:41:42
|
Hi, On Wed, 2002-10-30 at 20:42, Akos Balazs wrote: > Hi Gerrit, > > > > What's this substitute_header stuff? What I don't understand is that if > > > you have a minimalistic program, say: > [...] > > > > > > you'll get lotsof errors because of stdlib.h during compilation with icc. > > > That's why I thought it to be a gcc 3.2 vs icc 6.0 issue... > > > > that's because intel replaces the cdefs.h file which includes > > all the new namespace defines from gcc 3.2. That's I guess 90% of > > the error you see. > > > Oh, I get it now. I was still half asleep I guess... :) > > > > > Probably I'll send intel my patchset too ;-). > > > > > > > > As soon as the compilation is finished I'll post the headers. > > > > > > What patchset are you talking about? Do you mean you actually have to > > > patch the system-wide headers? Ouch... :\ > > > > > > > attached you find one tgz archive which contains a new substitute_header > > directory, just move the old one away and untar the archive and one > > patch for include/hash_map. Both only affects the intel/ia32/ > > installation, so no need to change the system. > > Thanks a lot, now it seems to compile. > > > hope it works, at least it compiles for me, now it's time to rebuild > > libGLU.so ;-) > > Could you send this to me aswell if you're building on RedHat 8.0 with > gcc? (I don't want my system-wide stuff to be compiled with icc, although > in the case of C sources it probably doesn't (and shouldn't) matter...) I will, currently 8.0 behaves like the typical .0 release you have to tweak it a little bit to work ;-). I already have the sources installed so as soon as I have the binary I will send out a copy. gerrit |
From: Gerrit V. <vo...@ca...> - 2002-10-31 03:57:52
Attachments:
GLU_rh_8.0_icc.tgz
|
Hi, ok, attached you find a set of lihGLU's for Redhat 8.0 compiled with the icc, using them testWindowGLUT runs. BTW in general where should we put stuff like this, probably I use one of the webservers overhere. gerrit On Thu, 2002-10-31 at 10:41, Gerrit Voss wrote: > > Hi, > > On Wed, 2002-10-30 at 20:42, Akos Balazs wrote: > > Hi Gerrit, > > > > > > What's this substitute_header stuff? What I don't understand is that if > > > > you have a minimalistic program, say: > > [...] > > > > > > > > you'll get lotsof errors because of stdlib.h during compilation with icc. > > > > That's why I thought it to be a gcc 3.2 vs icc 6.0 issue... > > > > > > that's because intel replaces the cdefs.h file which includes > > > all the new namespace defines from gcc 3.2. That's I guess 90% of > > > the error you see. > > > > > > Oh, I get it now. I was still half asleep I guess... :) > > > > > > > Probably I'll send intel my patchset too ;-). > > > > > > > > > > As soon as the compilation is finished I'll post the headers. > > > > > > > > What patchset are you talking about? Do you mean you actually have to > > > > patch the system-wide headers? Ouch... :\ > > > > > > > > > > attached you find one tgz archive which contains a new substitute_header > > > directory, just move the old one away and untar the archive and one > > > patch for include/hash_map. Both only affects the intel/ia32/ > > > installation, so no need to change the system. > > > > Thanks a lot, now it seems to compile. > > > > > hope it works, at least it compiles for me, now it's time to rebuild > > > libGLU.so ;-) > > > > Could you send this to me aswell if you're building on RedHat 8.0 with > > gcc? (I don't want my system-wide stuff to be compiled with icc, although > > in the case of C sources it probably doesn't (and shouldn't) matter...) > > I will, currently 8.0 behaves like the typical .0 release you have to > tweak it a little bit to work ;-). I already have the sources installed > so as soon as I have the binary I will send out a copy. > > gerrit > > > > > ------------------------------------------------------- > This sf.net email is sponsored by: Influence the future > of Java(TM) technology. Join the Java Community > Process(SM) (JCP(SM)) program now. > http://ads.sourceforge.net/cgi-bin/redirect.pl?sunm0004en > _______________________________________________ > Opensg-users mailing list > Ope...@li... > https://lists.sourceforge.net/lists/listinfo/opensg-users |
From: Gerrit V. <vo...@ca...> - 2002-10-31 04:11:59
|
Hi, ok this one got stuck, try http://155.69.99.20/user/gerrit/GLU_rh_8.0_icc.tgz gerrit On Thu, 2002-10-31 at 11:56, Gerrit Voss wrote: > > Hi, > > ok, attached you find a set of lihGLU's for Redhat 8.0 compiled with > the icc, using them testWindowGLUT runs. > > BTW in general where should we put stuff like this, probably I > use one of the webservers overhere. > > gerrit > > On Thu, 2002-10-31 at 10:41, Gerrit Voss wrote: > > > > Hi, > > > > On Wed, 2002-10-30 at 20:42, Akos Balazs wrote: > > > Hi Gerrit, > > > > > > > > What's this substitute_header stuff? What I don't understand is that if > > > > > you have a minimalistic program, say: > > > [...] > > > > > > > > > > you'll get lotsof errors because of stdlib.h during compilation with icc. > > > > > That's why I thought it to be a gcc 3.2 vs icc 6.0 issue... > > > > > > > > that's because intel replaces the cdefs.h file which includes > > > > all the new namespace defines from gcc 3.2. That's I guess 90% of > > > > the error you see. > > > > > > > > > Oh, I get it now. I was still half asleep I guess... :) > > > > > > > > > Probably I'll send intel my patchset too ;-). > > > > > > > > > > > > As soon as the compilation is finished I'll post the headers. > > > > > > > > > > What patchset are you talking about? Do you mean you actually have to > > > > > patch the system-wide headers? Ouch... :\ > > > > > > > > > > > > > attached you find one tgz archive which contains a new substitute_header > > > > directory, just move the old one away and untar the archive and one > > > > patch for include/hash_map. Both only affects the intel/ia32/ > > > > installation, so no need to change the system. > > > > > > Thanks a lot, now it seems to compile. > > > > > > > hope it works, at least it compiles for me, now it's time to rebuild > > > > libGLU.so ;-) > > > > > > Could you send this to me aswell if you're building on RedHat 8.0 with > > > gcc? (I don't want my system-wide stuff to be compiled with icc, although > > > in the case of C sources it probably doesn't (and shouldn't) matter...) > > > > I will, currently 8.0 behaves like the typical .0 release you have to > > tweak it a little bit to work ;-). I already have the sources installed > > so as soon as I have the binary I will send out a copy. > > > > gerrit > > > > > > > > > > ------------------------------------------------------- > > This sf.net email is sponsored by: Influence the future > > of Java(TM) technology. Join the Java Community > > Process(SM) (JCP(SM)) program now. > > http://ads.sourceforge.net/cgi-bin/redirect.pl?sunm0004en > > _______________________________________________ > > Opensg-users mailing list > > Ope...@li... > > https://lists.sourceforge.net/lists/listinfo/opensg-users > |
From: Akos B. <edh...@cs...> - 2002-10-31 14:46:06
|
Hi guys, Just wanted to share, I successfully compiled the CVS version of OpenSG on RedHat 8.0 with icc 6.0.1, even the tests compiled and run. :-) Thanks to Gerrit for all the patches here and there, and boo's to Intel :-) Akos Balazs |
From: Akos B. <edh...@cs...> - 2002-10-30 13:51:52
|
Hi, OK I know this is not the intel-compiler helpline, but the cheering was - unfortunately - too early, it still doesn't compile... :((( Sometimes it dies with segfault in 'mcpcom', I really have no idea why... :\ It's part of icc and gets invoked by icc somehow, but I don't know the details. A relevant buldlog entry is: make[2]: Entering directory `/local/work/OpenSG-CVS/OpenSG/Builds/i686-pc-linux-gnu-icc/BaseLib' /local/work/OpenSG-CVS/OpenSG/Common/commonBuildDetect.mk:43: Detected DBG /local/work/OpenSG-CVS/OpenSG/Common/commonBuildDetect.mk:59: Detected Shared Lib /local/work/OpenSG-CVS/OpenSG/Common/common.i686-pc-linux-gnu-icc.mk:125: Got Compiler Version : 600 using BD = /local/work/OpenSG-CVS/OpenSG/Source icc -w2 -wd981,424,76 -D_GNU_SOURCE -DQT_CLEAN_NAMESPACE -DOSG_WITH_GLUT -DOSG_WITH_QT -DOSG_WITH_JPG -DOSG_WITH_PNG -DOSG_WITH_MNG -DOSG_SUPPORT_NO_GEO_INTERFACE=1 -D_OSG_HAVE_CONFIGURED_H_ -D__INTEL_COMPILER_VERSION=600 -DQT_NO_XINERAMA -DQT_NO_XRENDER -DQT_NO_XFTFREETYPE -DQT_NO_XKB -DQT_NO_SM_SUPPORT -DQT_NO_IMAGEIO_MNG -DQT_NO_IMAGEIO_JPEG -DQT_NO_STYLE_AQUA -DQT_NO_STYLE_MAC -DQT_NO_STYLE_INTERLACE -DQT_NO_STYLE_COMPACT -Tnoauto -KPIC -Kc++eh -Krtti -ansi -O2 -mp -g -DOSG_DEBUG -inline_debug_info -wd981,424,76 -c -I"../Base" -I$BD/Base/Base -I$BD/Base/Field -I$BD/Base/Functors -I$BD/Base/Network/Base -I$BD/Base/Network/Socket -I$BD/Base/StringConversion \ -Iobj-dbg -I. \ -o obj-dbg/OSGBarrier.o /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGBarrier.cpp /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGBarrier.cpp icc: error: Fatal error in /opt/intel/compiler60/ia32/bin/mcpcom, terminated by segmentation violation compilation aborted for /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGBarrier.cpp (code 1) make[2]: *** [obj-dbg/OSGBarrier.o] Error 1 ... and I get lots of these. I can supply a complete buildlog, if necessary. May these be caused by the hacked headerfiles? Maybe I should go back to RedHat 7.0... :\ I also get some juicy errors like: icc -w2 -wd981,424,76 -D_GNU_SOURCE -DQT_CLEAN_NAMESPACE -DOSG_WITH_GLUT -DOSG_WITH_QT -DOSG_WITH_JPG -DOSG_WITH_PNG -DOSG_WITH_MNG -DOSG_SUPPORT_NO_GEO_INTERFACE=1 -D_OSG_HAVE_CONFIGURED_H_ -D__INTEL_COMPILER_VERSION=600 -DQT_NO_XINERAMA -DQT_NO_XRENDER -DQT_NO_XFTFREETYPE -DQT_NO_XKB -DQT_NO_SM_SUPPORT -DQT_NO_IMAGEIO_MNG -DQT_NO_IMAGEIO_JPEG -DQT_NO_STYLE_AQUA -DQT_NO_STYLE_MAC -DQT_NO_STYLE_INTERLACE -DQT_NO_STYLE_COMPACT -Tnoauto -KPIC -Kc++eh -Krtti -ansi -O2 -mp -g -DOSG_DEBUG -inline_debug_info -wd981,424,76 -c -I"../Base" -I$BD/Base/Base -I$BD/Base/Field -I$BD/Base/Functors -I$BD/Base/Network/Base -I$BD/Base/Network/Socket -I$BD/Base/StringConversion \ -Iobj-dbg -I. \ -o obj-dbg/OSGTypeFactory.o /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGTypeFactory.cpp /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGTypeFactory.cpp /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGTypeFactory.cpp(124): remark #383: value copied to temporary, reference to temporary used _vTypeNameMaps.push_back(new TypeNameMap); ^ /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGTypeFactory.cpp(232): remark #383: value copied to temporary, reference to temporary used _vTypeStore.push_back(NULL); ^ /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGTypeFactory.cpp(234): remark #383: value copied to temporary, reference to temporary used _vTypeNameMaps.push_back(new TypeNameMap); ^ /opt/intel/compiler60/ia32/include/map(140): error: identifier "insert" is undefined insert(value_type(_Keyval, mapped_type())).first; ^ detected during instantiation of "std::map<_Kty, _Ty, _Pr, _Alloc>::mapped_type &std::map<_Kty, _Ty, _Pr, _Alloc>::operator[](const std::map<_Kty, _Ty, _Pr, _Alloc>::key_type &) [with _Kty=osg::IDStringLink, _Ty=osg::UInt32={u_int32_t={unsigned int}}, _Pr=std::less<osg::IDStringLink>, _Alloc=std::allocator<std::pair<const osg::IDStringLink, osg::UInt32={u_int32_t={unsigned int}}>>]" icc: error: Fatal error in /opt/intel/compiler60/ia32/bin/mcpcom, terminated by segmentation violation compilation aborted for /local/work/OpenSG-CVS/OpenSG/Source/Base/Base/OSGTypeFactory.cpp (code 1) make[2]: *** [obj-dbg/OSGTypeFactory.o] Error 1 (See the part the complains about "insert" not being defined). having "fun" :), Akos |
From: Akos B. <edh...@cs...> - 2002-10-30 18:30:06
|
Hi guys, In case I haven't bored everyone (esp. Gerrit :) to death with my icc problems, here comes the next dose... :( I upgraded icc to 6.0.1, and OpenSG from CVS a few minutes ago. Unfortunately compilation still isn't possible: The segfaults are gone, so that was definitely a fault from the earlier build of icc, but I still have some strange errors - they seem to be related to intel-supplied headerfiles, but I don't think I'll be able to fick these... :( /local/work/OpenSG-CVS/OpenSG/Source/Base/Field/OSGBinaryDataHandler.inl(149): error: identifier "__bswap_32" is undefined UInt32 z = htonl(value); ^ /local/work/OpenSG-CVS/OpenSG/Source/Base/Field/OSGBinaryDataHandler.inl(169): error: expected an expression Int16 z = htons(value); ^ This is of course all over the place, where the BinaryDataHandler is used. :) This seems to be a problem with the network byte order conversion, I've followed which headerfiles are used for this but since I don't know the exact specification in which order icc tries to search for headerfiles I sorta gave up. :) /opt/intel/compiler60/ia32/include/hash_map(336): error: identifier "insert" is undefined insert(value_type(_Keyval, mapped_type())).first; ^ detected during instantiation of "std::hash_map<_Kty, _Ty, std::hash<_Kty>, _Keyeq>::mapped_type &std::hash_map<_Kty, _Ty, std::hash<_Kty>, _Keyeq>::operator[](const std::hash_map<_Kty, _Ty, std::hash<_Kty>, _Keyeq>::key_type &) [with _Kty=const osg::Char8={char} *, _Ty=osg::VRMLNodeDesc *, _Keyeq=osg::EQString]" This one definitely looks like a bug in the intel supplied hash_map STL headerfile... :\ Oh, and a lot of remarks like: "remark #193: zero used for undefined preprocessing identifier...", but I figure these are probably harmless. :) icc versions: $ rpm -qa|grep intel intel-isubh6-6.0.1-304 intel-ildb6-6.0.1-308 intel-icc6-6.0.1-304 Regards, Akos Balazs |
From: Gerrit V. <vo...@ca...> - 2002-10-31 02:29:30
|
Hi, On Thu, 2002-10-31 at 02:29, Akos Balazs wrote: > Hi guys, > > In case I haven't bored everyone (esp. Gerrit :) to death with my icc > problems, here comes the next dose... :( not yet ;-) > I upgraded icc to 6.0.1, and OpenSG from CVS a few minutes ago. > Unfortunately compilation still isn't possible: > > The segfaults are gone, so that was definitely a fault from the earlier > build of icc, but I still have some strange errors - they seem to be > related to intel-supplied headerfiles, but I don't think I'll be able to > fick these... :( > > /local/work/OpenSG-CVS/OpenSG/Source/Base/Field/OSGBinaryDataHandler.inl(149): error: identifier "__bswap_32" is undefined > UInt32 z = htonl(value); > ^ > > /local/work/OpenSG-CVS/OpenSG/Source/Base/Field/OSGBinaryDataHandler.inl(169): error: expected an expression > Int16 z = htons(value); > ^ hmm this one is new ;-), I'll try if vmware stops trashing my machine ;-). > This is of course all over the place, where the BinaryDataHandler is used. > :) This seems to be a problem with the network byte order conversion, I've > followed which headerfiles are used for this but since I don't know the > exact specification in which order icc tries to search for headerfiles I > sorta gave up. :) > > /opt/intel/compiler60/ia32/include/hash_map(336): error: identifier "insert" is undefined > insert(value_type(_Keyval, mapped_type())).first; > ^ > detected during instantiation of "std::hash_map<_Kty, _Ty, std::hash<_Kty>, _Keyeq>::mapped_type &std::hash_map<_Kty, _Ty, std::hash<_Kty>, _Keyeq>::operator[](const std::hash_map<_Kty, _Ty, std::hash<_Kty>, _Keyeq>::key_type &) [with _Kty=const osg::Char8={char} *, _Ty=osg::VRMLNodeDesc *, _Keyeq=osg::EQString]" > > This one definitely looks like a bug in the intel supplied hash_map STL > headerfile... :\ yes, but it should be fixed with the patch (hash_map.diff) I attached to my last mail, it is just some simple lookup problem. If the patch does not work for you just change insert(value_type ....) to this->insert(value_type ....) > Oh, and a lot of remarks like: "remark #193: zero used for undefined > preprocessing identifier...", but I figure these are probably harmless. :) yes they are, but I did not look closer at them so they are still in there to remind me ;-) regards gerrit > icc versions: > $ rpm -qa|grep intel > intel-isubh6-6.0.1-304 > intel-ildb6-6.0.1-308 > intel-icc6-6.0.1-304 > > > Regards, > Akos Balazs > > > > > > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > Opensg-users mailing list > Ope...@li... > https://lists.sourceforge.net/lists/listinfo/opensg-users |