MySQL's configure script greps /usr/include/pthread.h for "Linuxthreads" (configure line 12943, at least as of 4.0.15) to determine if Linux Threads are available. With an nptl-enabled glibc, Linuxthreads is _not_ in /usr/include/pthread.h, so this fails, and MySQL's configure script fails: ... checking "LinuxThreads"... "Not found" configure: error: This is a linux system and Linuxthreads was not found. On linux Linuxthreads should be used. Please install Linuxthreads (or a new glibc) and try again. See the Installation chapter in the Reference Manual for more information. !!! ERROR: dev-db/mysql-4.0.15 failed. !!! Function econf, Line 324, Exitcode 1 !!! econf failed So, what's needed to get MySQL to compile on a recent glibc+nptl+kernel system? My emerge info: Portage 2.0.49-r4 (default-x86-1.4, gcc-3.3.1, glibc-2.3.2-r5, 2.6.0-test4-mm2) ================================================================= System uname: 2.6.0-test4-mm2 i686 Intel(R) Pentium(R) M processor 1400MHz distcc 2.10 i686-pc-linux-gnu (protocols 1 and 2) (default port 3632) [enabled] ACCEPT_KEYWORDS="x86 ~x86 amd64" AUTOCLEAN="yes" CFLAGS="-march=pentium4 -O3 -mmmx -msse -msse2 -mfpmath=sse -pipe -fomit-frame-pointer -funroll-loops -frerun-loop-opt" CHOST="i686-pc-linux-gnu" COMPILER="gcc3" CONFIG_PROTECT="/etc /var/qmail/control /usr/share/config /usr/kde/2/share/config /usr/kde/3/share/config /usr/X11R6/lib/X11/xkb" CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d" CXXFLAGS="-march=pentium4 -O3 -mmmx -msse -msse2 -mfpmath=sse -pipe -fomit-frame-pointer -funroll-loops -frerun-loop-opt" DISTDIR="/usr/portage/distfiles" FEATURES="sandbox ccache autoaddcvs distcc" GENTOO_MIRRORS="ftp://gentoo.mirrors.pair.com" MAKEOPTS="-j3" PKGDIR="/usr/portage/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/usr/portage" PORTDIR_OVERLAY="/usr/local/portage" SYNC="rsync://backup/gentoo-portage" USE="x86 oss avi crypt cups encode foomaticdb gif jpeg libwww mad mikmod mmx mpeg ncurses pdflib png quicktime spell truetype xml2 xmms xv zlib directfb gtkhtml alsa gdbm berkdb slang readline aalib bonobo tcltk guile mysql X sdl gpm tcpd pam ssl perl imlib oggvorbis gnome gtk opengl mozilla cdr aavm acpi acpi4linux dga dvd fbcon gd gnomedb gtk2 icc mbox moznoirc nptl pcmcia pic prebuilt samba sse sse2 threads tiff trusted usb vim-with-x wmf xosd -3dnow -apm -arts -java -kde -libg++ -motif -nls -qt -svga -python -esd"
Incidentally, this also happened with glibc-2.3.2-r3 - I upgraded to -r5 to see if it would change anything. Someone reported it on the forums with -r1 as well (http://forums.gentoo.org/viewtopic.php?t=86240).
Same problem with me: glibc 2.3.2-r1 and 2.3.2-r5 Message: checking for TCP wrappers library -lwrap... yes checking for atomic operations... atomic_add atomic_sub checking if we should use pstack... no checking for int8... no checking "LinuxThreads"... "Not found" configure: error: This is a linux system and Linuxthreads was not found. On linux Linuxthreads should be used. Please install Linuxthreads (or a new glibc) and try again. See the Installation chapter in the Reference Manual for more information. !!! ERROR: dev-db/mysql-4.0.15 failed. !!! Function econf, Line 324, Exitcode 1 !!! econf failed emerge info gentoo root # emerge info Portage 2.0.49-r4 (default-x86-1.4, gcc-3.3.1, glibc-2.3.2-r1, 2.6.0-test5-love3) ================================================================= System uname: 2.6.0-test5-love3 i686 AMD Athlon(tm) XP 2000+ distcc 2.10 i686-pc-linux-gnu (protocols 1 and 2) (default port 3632) [disabled] ccache version 2.2 [enabled] ACCEPT_KEYWORDS="x86 ~x86" AUTOCLEAN="yes" CFLAGS="-march=athlon-xp -m3dnow -msse -mfpmath=sse -mmmx -O3 -pipe -fforce-addr -fomit-frame-pointer -funroll-loops -frerun-cse-after-loop -frerun-loop-opt -falign-functions=4 -maccumulate-outgoing-args -ffast-math -fprefetch-loop-arrays" CHOST="i686-pc-linux-gnu" COMPILER="gcc3" CONFIG_PROTECT="/etc /var/qmail/control /usr/kde/2/share/config /usr/kde/3/share/config /usr/X11R6/lib/X11/xkb /usr/share/texmf/tex/generic/config/ /usr/share/texmf/tex/platex/config/ /usr/share/config" CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d" CXXFLAGS="-march=athlon-xp -m3dnow -msse -mfpmath=sse -mmmx -O3 -pipe -fforce-addr -fomit-frame-pointer -funroll-loops -frerun-cse-after-loop -frerun-loop-opt -falign-functions=4 -maccumulate-outgoing-args -ffast-math -fprefetch-loop-arrays" DISTDIR="/software/distfiles" FEATURES="autoaddcvs -sandbox ccache" GENTOO_MIRRORS="http://gentoo.oregonstate.edu http://distro.ibiblio.org/pub/Linux/distributions/gentoo" MAKEOPTS="-j2" PKGDIR="/usr/portage/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/usr/portage" PORTDIR_OVERLAY="/usr/local/portage" SYNC="rsync://rsync.gentoo.org/gentoo-portage" USE="x86 oss apm avi crypt encode foomaticdb gif jpeg libg++ mad mikmod mmx mpeg ncurses pdflib png quicktime truetype xml2 xmms xv zlib gtkhtml gdbm berkdb slang readline tetex bonobo svga tcltk java guile mysql X sdl gpm tcpd pam libwww ssl perl python esd imlib oggvorbis gnome motif opengl mozilla dvd radeon sse 3dnow gtk gtk2 dga -qt -kde -arts -spell -alsa breakme pda nls lirc cups nptl postgres"
Any ideas?
I don't know if this helps and/or fixes the current problem, and for now, I only see it as a hack and not as a solution. I don't even know if mysql will start and work properly, since I am compiling my new system (2.6-test5-love3 with nptl) from scratch. But at least with this patch to the mysql ebuild it compiled on my system: --- schnipp --- /var/portage/dev-db/mysql/mysql-4.0.15.ebuild 2003-09-24 22:11:48.000000000 +0200 +++ /mnt/gentoo/var/overportage/dev-db/mysql/mysql-4.0.15.ebuild 2003-09-30 22:00:15.000000000 +0200 @@ -21,7 +21,7 @@ KEYWORDS="ia64 ~x86 ~sparc ~ppc ~arm ~amd64 ~hppa" LICENSE="GPL-2" SLOT="0" -IUSE="static readline innodb berkdb tcpd ssl perl debug" +IUSE="static readline innodb berkdb tcpd ssl perl debug nptl" DEPEND="readline? ( >=sys-libs/readline-4.1 ) tcpd? ( >=sys-apps/tcp-wrappers-7.6-r6 ) @@ -103,6 +103,9 @@ myconf="${myconf} `use_with tcpd libwrap`" myconf="${myconf} `use_with innodb`" + use nptl && myconf="${myconf} --with-pthread --with-named-thread-libs=-lpthread" + use nptl && myconf="${myconf} --with-pthread=no" + use ssl \ && myconf="${myconf} --with-vio --with-openssl" \ || myconf="${myconf} --without-openssl" @@ -136,6 +139,15 @@ --with-embedded-server \ ${myconf} || die "bad ./configure" + if use nptl; then + echo -n "#undef HAVE_PTHREAD_RWLOCK_RDLOCK" >>config.h + echo -n "#define USE_MUTEX_INSTEAD_OF_RW_LOCKS 1" >>config.h +# echo -n "#undef THREAD" >>config.h + sed -s "s/ADAPTIVE/REMOVEDCRAP/" include/my_pthread.h >asdf + cp asdf include/my_pthread.h + echo -n "#undef PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP" >>include/my_pthread.h + fi + emake || die "compile problem" } --- schnapp The changes are the following: 1. I have explicitly stated that pthread is the thread library (and not LinuxThreads) 2. there are some problems with the read/write locks from nptl - don't know whose fault it is mysql's or nptl's -> using mutexes works 3. adaptive mutexes breaks the linker (can't find a certain function)
If I'm not mistaken, shouldn't the second line be || instead of &&? + use nptl && myconf="${myconf} --with-pthread --with-named-thread-libs=-lpthread" + use nptl && myconf="${myconf} --with-pthread=no"
Oops! No, this line should be removed, my mistake. Maybe it did work for me because --with-pthread will be ignored when you use --with-named-thread-libs.
Your patch doesn't do what you think it does. For some reason, you have -n's on all your echos, so everything ends up on one line at the end of config.h: #undef HAVE_PTHREAD_RWLOCK_RDLOCK#define USE_MUTEX_INSTEAD_OF_RW_LOCKS 1 Additionally, the lack of newlines produces warnings: ../../include/my_config.h:871:73: warning: no newline at end of file ../../include/my_pthread.h:686:45: warning: no newline at end of file But, when I changed them to echo's without -n's, compilation fails with a series of syntax errors in thr_rwlock.c. I uncommented the 'echo "#undef THREADS"' (and removed the -n), and it then gets much further, but still dies with: gcc -O3 -DDBUG_OFF -march=pentium4 -O3 -mmmx -msse -msse2 -mfpmath=sse -pipe -fomit-frame-pointer -funroll-loops -frerun-loop-opt -DHAVE_ERRNO_AS_DEFINE=1 -DUSE_OLD_FUNCTIONS -rdynamic -o test-sslclient test-sslclient.o -lstdc++ ../dbug/libdbug.a libvio.a ../mysys/libmysys.a ../strings/libmystrings.a -L/usr/lib -lssl -lcrypto -lpthread -lz -lcrypt -lnsl -lm -lpthread test-sslserver.o(.text+0x281): In function `client_thread': : undefined reference to `my_thread_init' collect2: ld returned 1 exit status make[2]: *** [test-sslserver] Error 1 make[2]: *** Waiting for unfinished jobs.... make[2]: Leaving directory `/var/tmp/portage/mysql-4.0.16/work/mysql-4.0.16/vio' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory `/var/tmp/portage/mysql-4.0.16/work/mysql-4.0.16' make: *** [all] Error 2 !!! ERROR: dev-db/mysql-4.0.16 failed. !!! Function src_compile, Line 153, Exitcode 2 !!! compile problem I should note that the original patch (which really wasn't doing anything significant to config.h, but _was_ changing my_pthread.h) compiled MySQL successfully, and appeared to run properly, though I only ran a few simple queries. So how important are these lines: #undef HAVE_PTHREAD_RWLOCK_RDLOCK #define USE_MUTEX_INSTEAD_OF_RW_LOCKS 1 Without them, it compiles, and appears to run; with them, it fails.
*** Bug 32240 has been marked as a duplicate of this bug. ***
I have managed to get mysql to compile properly by adding '--with-named-thread-libs=-pthread'. To do this from the ebuild script, add use nptl && export CFLAGS="${CFLAGS} --with-named-thread-libs=-pthread" somewhere before the econf line, and nptl to IUSE. I'll see if I can make a diff, if that isn't clear enough :)
um. I made a big typo there. It's myconf that needs to be set - use nptl && export myconf="${myconf} --with-named-thread-libs=-pthread"
never mind - this is part of the patch above, and has some problems. someone bash me over the head for not checking the other patch fully.
I've had success with both 4.0.16 and 4.1.0 by changing configure to automatically assume LinuxThreads are "found", whether or not they actually are. The two ways to do this seem to be: a) Add "/* Linuxthreads */" into /usr/include/pthreads.h, or b) Change configure to just assume that LinuxThreads are there. Neither solution is very nice, and MySQL's detection of "grep LinuxThreads" is just plain idiotic. However, it does compile, and does work with either of these solutions (and the fact that MySQL finally includes subselect support in 4.1.0 is pretty neat). The MySQL team itself won't help on this issue - it works on Redhat with their hacked in 2.4 NPTL support, so they seem to be refusing to fix any bugs - could the "fix" that Redhat has be so simple as just adding /* Linuxthreads */ into pthreads.h?
Re: Comment #12: Jason, I like method (a)! I edited /usr/include/pthread.h by hand, but a patch could edit it and then restore the original after the compile, if necessary.
Created attachment 21563 [details, diff] mysql-nptl.patch Modifies configure.in to pretend we really have LinuxThreads, even when we don't.
Created attachment 21565 [details, diff] NPTL ebuild patch * Add IUSE="nptl". * Add DEPEND="nptl? ( sys-devel-autoconf )" * If USE="nptl", apply mysql-nptl.patch and run autoconf. This compiles successfully and seems to work but has not been thoroughly tested.
Patch from above comment works perfectly. Please include in official ebuild
*** Bug 35999 has been marked as a duplicate of this bug. ***
Created attachment 22649 [details, diff] MySQL 4.0.16 ebuild, supports NPTL There is a VERY CLEAN fix for this... add use nptl && myconf="${myconf} --with-named-thread-libs=-lpthread " then, use "-D_GNU_SOURCE" in CPPFLAGS. I've checked that it works. It (rightfully) avoids using LinuxThreads, because NPTL is not LinuxThreads.
Still can't get mysql to compile MySQL 4.0.16 with the customized ebuild. It starts to compile but after a while the following errors (or very similar to it are repeated several times gcc -DMAP_TO_USE_RAID -I. -I. -I.. -I./../include -I../include -O3 -DDBUG_OFF -O2 -mcpu=pentium4 -funroll-loops -fomit-frame-pointer -pipe -DHAVE_ERRNO_AS_DEFINE=1 -DUSE_OLD_FUNCTIONS -c `test -f mi_rnext.c || echo './'`mi_rnext.c In file included from mi_extra.c:17: myisamdef.h:206: error: syntax error before "pthread_rwlock_t" myisamdef.h:206: warning: no semicolon at end of struct or union myisamdef.h:208: warning: data definition has no type or storage class myisamdef.h:221: error: syntax error before "MYISAM_SHARE" myisamdef.h:221: warning: no semicolon at end of struct or union myisamdef.h:246: error: `checksum' redeclared as different kind of symbol ../include/my_sys.h:753: error: previous declaration of `checksum' myisamdef.h:274: error: syntax error before '}' token myisamdef.h:492: error: syntax error before '*' token In file included from mi_info.c:19: myisamdef.h:206: error: syntax error before "pthread_rwlock_t" myisamdef.h:206: warning: no semicolon at end of struct or union myisamdef.h:208: warning: data definition has no type or storage class myisamdef.h:221: error: syntax error before "MYISAM_SHARE" myisamdef.h:221: warning: no semicolon at end of struct or union myisamdef.h:246: error: `checksum' redeclared as different kind of symbol ../include/my_sys.h:753: error: previous declaration of `checksum' myisamdef.h:274: error: syntax error before '}' token myisamdef.h:660: error: syntax error before "MYISAM_SHARE" myisamdef.h:661: error: syntax error before '*' token myisamdef.h:662: error: syntax error before '*' token myisamdef.h:492: error: syntax error before '*' token System info: Portage 2.0.49-r15 (default-x86-1.4, gcc-3.3.2, glibc-2.3.2-r9, 2.6.0) ================================================================= System uname: 2.6.0 i686 Intel(R) Xeon(TM) CPU 2.66GHz Gentoo Base System version 1.4.3.10 ACCEPT_KEYWORDS="x86" AUTOCLEAN="yes" CFLAGS="-O2 -mcpu=pentium4 -funroll-loops -fomit-frame-pointer -pipe" CHOST="i686-pc-linux-gnu" COMPILER="gcc3" CONFIG_PROTECT="/etc /var/qmail/control /usr/share/config /usr/kde/2/share/config /usr/kde/3/share/config" CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d" CXXFLAGS="-O2 -mcpu=pentium4 -funroll-loops -fomit-frame-pointer -pipe" DISTDIR="/usr/portage/distfiles" FEATURES="sandbox ccache autoaddcvs" GENTOO_MIRRORS="http://trumpetti.atm.tut.fi/gentoo/" MAKEOPTS="-j5" PKGDIR="/usr/portage/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/usr/portage" PORTDIR_OVERLAY="/usr/local/portage" SYNC="rsync://rsync.europe.gentoo.org/gentoo-portage" USE="x86 crypt cups encode foomaticdb gif gpm imlib jpeg libg++ mad mikmod mpeg ncurses nls pdflib png svga truetype xml2 xv zlib gdbm berkdb slang readline mysql tcpd pam libwww ssl perl python acpi acpi4linux apache2 gd imap innodb ipv6 java kerberos krb4 mmx noaudio nptl pic regexp sse tiff -oss -apm -arts -avi -gtk -gtk2 -kde -gnome -motif -oggvorbis -opengl -qt -quicktime -sdl -spell -X -xmms" I also tried 4.0.17, but it didn't work either.
Created attachment 22942 [details, diff] Mysql 4.0.17 nptl aware ebuild Based on the build from Zhen Lin: tell mysql eplicitely about the thread library and then configure with -D_GNU_SOURCE in CPPFLAGS. I had to prefix the CPPFLAGS to the configure command, else they weren't used by configure. This caused a build error much like Tony Soukas shows.
Created attachment 23034 [details, diff] Updated nptl ebuild patch This patch is based on the work in comment #20 and comment #18. There are some minor spacing cleanups and nptl is added to IUSE. It seems to pass basic smoke tests.
i just used the ebuild marked as 'Mysql 4.0.17 nptl aware ebuild' on 4.0.16 and after making the patch lines right for .16 it worked great
'basic smoke test' as your testing or the mysql offical tests? if it passes those, i'll be happy to commit it to the tree (and stop people bugging me about this). also, while you are at it for the tests, look very carefully at the memory usage, upstream mysql does NOT support NPTL at all, and many of these hacks like yours lead to massive memory usage. see http://bugs.mysql.com/bug.php?id=972
Adding self to CC. About the CPPFLAGS issue -- you have to export it in order for configure to pick it up, or specify it as an environment modifier. I've been running my modified MySQL since Christmas, no problems so far, RSS is 5.6MB. But, I don't have a 500-strong client load on it, so I can't really say.
The ebuild on comment #21 works. I haven't been able to test mysql in production use yet (probably some time next week), but I've run all the tests that come with mysql (except the innodb tests) and i've also run the the benchmarks that come bundled (although they seem to stress more perl than mysql ;). I've alse hammered the database with actual web server usage with ab2 and opensta, no problems so far.
*** Bug 39209 has been marked as a duplicate of this bug. ***
ok, this has been a bug for way too long, I use NPTL on my system, and have not been able to compile mysql, and it keeps blocking my emerge. I've committed a patch, it does not depend on USE flags, and it's meant only for gentoo systems, so do not send upstream. It should hit the rsync mirrors sometime today.
