PHP 5.3 won't compile on mips64
Hello everyone! I'm trying to compile PHP 5.3 on my mips64 (sgi) machine, because I not found this prebuilt package in 5.0 or snapshot repository. In file included from /usr/ports/pobj/php-5.3.10/php-5.3.10/sapi/fpm/fpm/fpm_scoreboard.h:15, from /usr/ports/pobj/php-5.3.10/php-5.3.10/sapi/fpm/fpm/fpm.c:21: /usr/ports/pobj/php-5.3.10/php-5.3.10/sapi/fpm/fpm/fpm_atomic.h:142:2: error: #error Unsupported processor. Please open a bug report (bugs.php.net). In file included from /usr/ports/pobj/php-5.3.10/php-5.3.10/sapi/fpm/fpm/fpm_scoreboard.h:15, from /usr/ports/pobj/php-5.3.10/php-5.3.10/sapi/fpm/fpm/fpm.c:21: /usr/ports/pobj/php-5.3.10/php-5.3.10/sapi/fpm/fpm/fpm_atomic.h:146: error: expected ')' before '*' token In file included from /usr/ports/pobj/php-5.3.10/php-5.3.10/sapi/fpm/fpm/fpm.c:21: /usr/ports/pobj/php-5.3.10/php-5.3.10/sapi/fpm/fpm/fpm_scoreboard.h:22: error: expected specifier-qualifier-list before 'atomic_t' /usr/ports/pobj/php-5.3.10/php-5.3.10/sapi/fpm/fpm/fpm_scoreboard.h:51: error: expected specifier-qualifier-list before 'atomic_t' *** Error code 1 Stop in /usr/ports/pobj/php-5.3.10/php-5.3.10 (line 560 of Makefile). *** Error code 1 I found the same bug and solution for ARM on bugs.php.net. Should I open a ticket about this bug on bugs.php.net or the problem can be solved right here? Unfortunately, I do not have the experience and knowledge in such low-level things. Anyway, for now, this port should be marked as broken at least for mips64. I do not think that anyone would be happy to see the same error after a couple of hours compiling :) Also, doxygen port in this platform can compile only by removing O2 flag or adding "--english-only" option to configure flags, but as far as I remember someone has already reported on this issue. Best regards, Val.
Re: gettext and Imagemagick packaging errors
I've tried.I wrote about this above. Without any mk.conf file. It was the first that occurred to me, although a bit and confused me. All option by default i.e. "clean" install with ports for 4.9 from ftp. Yes, I've installed gettext and ImageMagick from packages, but it would be bad if this error will exist out version to version. That's why I'm writing to a ports mailing list. I want find out what went wrong and how to build this package openbsd maintainers. 13 мая 2011 г. 4:54 пользователь Amit Kulkarni написал: > 2011/5/12 Валерий Немкович : >> I tried. The same. "-mips4" is not a tweak. This is an indication to >> the compiler to use the additional CPU instruction set and resulting >> software will work with fpu more effectively and faster. It is >> absolutely safe and stable, it is not aggresive flags for optimisation >> such as -O3, -funroll-all-loops or -ftree-vectorize. Another "big" >> ports such as apache2, postgresql, ruby interpreter were compiled and >> works fine. In any case any errors with wrong or non-optimal flags of >> compiler should be detected on compiling or linking stage, but not a >> package building. >> >> I don't have extensive experience in writing scripts for the ports >> and, therefore, ask for help or hints. The more so because I think it >> is a problem only applies to this platform ( mips64/sgi ) as this is a >> very popular ports and could not have been unnoticed before. >> >>>Obviously you should see if you can repeat this without your >>>mk.conf tweaks... >> > > You will not get any help if you use -mips4. It might be innocent but > gcc is a unpredictable beast. Try without it and report if any > problem. gettext exists for mips64, here's the proof > ftp://ftp.openbsd.org/pub/OpenBSD/4.9/packages/mips64/gettext-0.18.1p0.tgz > > This was unnoticed before and they ***might*** use that option in > future but right now, its unsupported. > > Good luck! >
Re: gettext and Imagemagick packaging errors
I installed gettext and Imagemagick from packages immediately once confronted with this errors. Before that I didn't use packages. It's just a habit and also an opportunity to get a 20% performance on some tasks (openvpn + lzo in my case). This is the standard compiler option, and does not affect the stability of the system. I've compiled about two dozen programs on OpenBSD ( and IRIX before ) with this option and never experienced problems in the compiling or stability. Again, I repeat, it is not aggressive optimization that modifies the intermediate code, such as the unrolling of loops, for example, that may affect the stability or stop the compiling due error at all. In any case, with "-mips" or not something is going wrong. I'm trying to find out the reason. I devote a couple of hours reading the documentation of creating port scripts, of course. I just thought that more familar with ports people will help me to more specifically isolate the problem. I will be grateful for any help. Best Regards, Val, hist...@gmail.com 2011/5/13 Amit Kulkarni : >> Yes, I've installed gettext and ImageMagick from packages, but it > > then why you install from ports? ports and packages is supposed to be same. > > do a make clean > > pkg_info | grep gettext > > If gettext is already installed and exists as a package in > /usr/ports/packages/SGI-MIPS64/all etc...? It will clash, see if you > can afford to pkg_delete and use ports to install gettext/Imagemagick. > > Look here for more info about ports > http://openbsd.org/faq/ports/testing.html >
Re: gettext and Imagemagick packaging errors
I've tried.I wrote about this above. Without any mk.conf file. It was the first that occurred to me, although a bit and confused me. All option by default i.e. "clean" install with ports for 4.9 from ftp. Yes, I've installed gettext and ImageMagick from packages, but it would be bad if this error will exist out version to version. That's why I'm writing to a ports mailing list. I want find out what went wrong and how to build this package openbsd maintainers. Best Regars, Val, hist...@gmail.com 13 мая 2011 г. 4:54 пользователь Amit Kulkarni написал: > 2011/5/12 Валерий Немкович : >> I tried. The same. "-mips4" is not a tweak. This is an indication to >> the compiler to use the additional CPU instruction set and resulting >> software will work with fpu more effectively and faster. It is >> absolutely safe and stable, it is not aggresive flags for optimisation >> such as -O3, -funroll-all-loops or -ftree-vectorize. Another "big" >> ports such as apache2, postgresql, ruby interpreter were compiled and >> works fine. In any case any errors with wrong or non-optimal flags of >> compiler should be detected on compiling or linking stage, but not a >> package building. >> >> I don't have extensive experience in writing scripts for the ports >> and, therefore, ask for help or hints. The more so because I think it >> is a problem only applies to this platform ( mips64/sgi ) as this is a >> very popular ports and could not have been unnoticed before. >> >>>Obviously you should see if you can repeat this without your >>>mk.conf tweaks... >> > > You will not get any help if you use -mips4. It might be innocent but > gcc is a unpredictable beast. Try without it and report if any > problem. gettext exists for mips64, here's the proof > ftp://ftp.openbsd.org/pub/OpenBSD/4.9/packages/mips64/gettext-0.18.1p0.tgz > > This was unnoticed before and they ***might*** use that option in > future but right now, its unsupported. > > Good luck! >
Re: gettext and Imagemagick packaging errors
I installed gettext and Imagemagick from packages immediately once confronted with this errors. Before that I didn't use packages. It's just a habit and also an opportunity to get a 20% performance on some tasks (openvpn + lzo in my case). This is the standard compiler option, and does not affect the stability of the system. I've compiled about two dozen programs on OpenBSD ( and IRIX before ) with this option and never experienced problems in the compiling or stability. Again, I repeat, it is not aggressive optimization that modifies the intermediate code, such as the unrolling of loops, for example, that may affect the stability or stop the compiling due error at all. In any case, with "-mips" or not something is going wrong. I'm trying to find out the reason. I devote a couple of hours reading the documentation of creating port scripts, of course. I just thought that more familar with ports people will help me to more specifically isolate the problem. I will be grateful for any help. Best Regards, Val, hist...@gmail.com 2011/5/13 Amit Kulkarni : >> Yes, I've installed gettext and ImageMagick from packages, but it > > then why you install from ports? ports and packages is supposed to be same. > > do a make clean > > pkg_info | grep gettext > > If gettext is already installed and exists as a package in > /usr/ports/packages/SGI-MIPS64/all etc...? It will clash, see if you > can afford to pkg_delete and use ports to install gettext/Imagemagick. > > Look here for more info about ports > http://openbsd.org/faq/ports/testing.html >
Re: gettext and Imagemagick packaging errors
Nothing changes. When I deleted mk.conf, and did "make clean=depends" and "make install" - I'm getting the same errors. I wrote about above. 2011/5/13 Stuart Henderson : > > On Fri, 13 May 2011 02:26:11 +0100, hist...@gmail.com wrote: >> I tried. > > Did it help? > > These packages build ok in bulk builds, so I think this is probably a local > problem. > >> The same. "-mips4" is not a tweak. This is an indication to >> the compiler to use the additional CPU instruction set and resulting >> software will work with fpu more effectively and faster. > > Similar to things like -march=k8 on amd64/i386. It's not tested and supported > (it may work but you're responsible for tracking down problems that occur as > a result). > > You can also make your machine work more effectively by using packages rather > than building from ports, this saves a huge amount of cpu _and_ human time... >
Re: C++ compiler on 4.9 is very unstable
2011/5/13 Miod Vallat : >> Very often when I try to compile anything from ports using G++, a >> compiler give me a "Segmentation Fault" or "Bus error". GCC works >> fine. And everytime I have to start from last stage. What can you >> advise? This happens on SGI O2, RM7000, 512MB of RAM. IRIX 6.5 using >> MIPSpro 7.4 can compile a few days without error so I think that the >> problem lies not in hardware. This is a bug in GCC 4.2.1, or just a >> particular build of 4.9? How do I fix this? Or is this the only >> problem I have one? > > Does this only happen when working on large c++ files? If so, have you > tried building with larger resources limits? The c++ compiler is > unfortunately very good at allocating a lot of memory, and very bad at > handling memory allocation failure. > > Miod > This happens on almost any c+ + port. Unfortunately, it is impossible to predict exactly when it happens and to reproduce the error. For example, when I tried to build textproc/icu4c ( it is dependence of PHP) I received the following errors. In the last example, I added " -v" in the Makefile so that the result was more evident ( and -save-temps also to obtain *.ii preprocessed source files). I had about forty times to resume the process of compilation, but then I got bored. I also added "-mips4" flag in /etc /mk.conf to improve perfomance with fpu in openvpn+lzo pair, but removal of this file and "make clean=depends" had no effect on the problem. Any suggestions? I will be grateful for any help. But someone, in the end, build those packages! Or packages in ports are result of cross-compiling? In any case, thanks to all. You guys are doing a truly titanic work. gmake[1]: Entering directory `/usr/ports/pobj/icu4c-4.4/build-mips64/i18n' c++ -D_REENTRANT -I../common -I/usr/ports/pobj/icu4c-4.4/icu/source/i18n -I/usr/ports/pobj/icu4c-4.4/icu/source/common -DU_I18N_IMPLEMENTATION -O2 -pipe -mips4 -W -Wall -ansi -pedantic -Wpointer-arith -Wwrite-strings -Wno-long-long -fvisibility=hidden -c -o nfrule.ao /usr/ports/pobj/icu4c-4.4/icu/source/i18n/nfrule.cpp In file included from /usr//include/g++/memory:60, from /usr//include/g++/string:48, from /usr/ports/pobj/icu4c-4.4/icu/source/common/unicode/std_string.h:39, from /usr/ports/pobj/icu4c-4.4/icu/source/common/unicode/unistr.h:31, from /usr/ports/pobj/icu4c-4.4/icu/source/common/unicode/strenum.h:14, from /usr/ports/pobj/icu4c-4.4/icu/source/common/unicode/uenum.h:24, from /usr/ports/pobj/icu4c-4.4/icu/source/common/unicode/uloc.h:25, from /usr/ports/pobj/icu4c-4.4/icu/source/i18n/unicode/ucol.h:18, from /usr/ports/pobj/icu4c-4.4/icu/source/i18n/unicode/coll.h:57, from /usr/ports/pobj/icu4c-4.4/icu/source/i18n/unicode/rbnf.h:30, from /usr/ports/pobj/icu4c-4.4/icu/source/i18n/nfrule.h:11, from /usr/ports/pobj/icu4c-4.4/icu/source/i18n/nfrule.cpp:16: /usr//include/g++/limits: In static member function 'static char std::numeric_limits::min()': /usr//include/g++/limits:375: warning: overflow in implicit constant conversion /usr//include/g++/limits: In static member function 'static wchar_t std::numeric_limits::max()': /usr//include/g++/limits:530: warning: overflow in implicit constant conversion /usr/ports/pobj/icu4c-4.4/icu/source/i18n/nfrule.cpp: In member function 'UBool icu::NFRule::allIgnorable(const icu::UnicodeString&, UErrorCode&) const': /usr/ports/pobj/icu4c-4.4/icu/source/i18n/nfrule.cpp:1431: internal compiler error: Segmentation fault Please submit a full bug report, with preprocessed source if appropriate. See http://gcc.gnu.org/bugs.html> for instructions. gmake[1]: *** [nfrule.ao] Error 1 gmake[1]: Leaving directory `/usr/ports/pobj/icu4c-4.4/build-mips64/i18n' gmake: *** [all-recursive] Error 2 *** Error code 2 Stop in /usr/ports/textproc/icu4c (line 2473 of /usr/ports/infrastructure/mk/bsd.port.mk). *** Error code 1 Stop in /usr/ports/textproc/icu4c (line 1690 of /usr/ports/infrastructure/mk/bsd.port.mk). *** Error code 1 Stop in /usr/ports/textproc/icu4c (line 2261 of /usr/ports/infrastructure/mk/bsd.port.mk). *** Error code 1 Stop in /usr/ports/textproc/icu4c (line 2241 of /usr/ports/infrastructure/mk/bsd.port.mk). *** Error code 1 Stop in /usr/ports/textproc/icu4c (line 1721 of /usr/ports/infrastructure/mk/bsd.port.mk). *** Error code 1 Stop in /usr/ports/textproc/icu4c (line 2241 of /usr/ports/infrastructure/mk/bsd.port.mk). *** Error code 1 Stop in /usr/ports/databases/openldap (line 1879 of /usr/ports/infrastructure/mk/bsd.port.mk). *** Error code 1 Stop in /usr/ports/databases/openldap (line 2293 of /usr/ports/infrastructure/mk/bsd.port.mk). *** Error code 1 Stop in /usr/ports/databases/openldap (line 1690 of /usr/ports/infrastructure/mk/bsd.port.mk). *** Error code 1 Stop in /usr/ports/databases/openldap (line 2261 of /usr/ports/infrastructure/mk/bsd.por