|
serg,
Do we still pretend on supporting Solaris Sparc?
|
|
I see "MDEV-5268 cannot build on Solaris 11 x86" I try it but some troubles. So I create this jira.
Pascal
|
|
and https://mariadb.com/kb/en/mariadb/buildbot-setup-for-solaris-sparc/
???
|
|
We don't have a Sparc builder at the moment. If you have a fix for this issue, we'll be happy to apply it. Otherwise it'll have to wait until we get a Sparc builder.
|
|
strings-t.dir/strings-t.c contents : my_sys.h
include/my_sys.h contents : my_pthread.h
include\my_pthread.h contents : extern int rw_pr_rdlock(rw_pr_lock_t *); @ 567
and in CMakeLists.txt we have :
@ line 377
INCLUDE(maintainer)
IF(WITH_UNIT_TESTS)
ENABLE_TESTING()
ADD_SUBDIRECTORY(unittest/mytap)
ADD_SUBDIRECTORY(unittest/strings) *breaks*
compile unittest/strings-t.c
and after
@ line 401 : ADD_SUBDIRECTORY(mysys)
|
|
Are you trying to describe a fix? I don't understand what we need to change, sorry. Could you please create a patch (unified diff) and attach it to this bug report? Thanks!
|
|
Hello Sergei,
I do not forget you.
I have troubles to go up a server Zope on Sparc 10 Solaris (openCSW , python and certificates).
So I decide to stop for a moment Solaris 10.
I had health worries, not much time of free. I bought a x5120 to install Solaris x86.
and I try to open the internet connection.
Pascal.
----- Mail original -----
De: "Sergei Golubchik (JIRA)" <jira@mariadb.atlassian.net>
À: pviltard@free.fr
Envoyé: Samedi 12 Décembre 2015 13:45:00
Objet: [JIRA] (MDEV-8448) cannot build on solaris 10 Sparc
[ https://mariadb.atlassian.net/browse/MDEV-8448?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sergei Golubchik updated MDEV-8448:
-----------------------------------
Priority: Major (was: Minor)
–
This message was sent by Atlassian JIRA
(v7.1.0-OD-02-030#71001)
|
|
No-no, don't worry. I've increased the priority because we've got a set of contributed patches for fixing compilation issues on Sparc. The priority is a reminder for me, not for you. And we do have a Sparc builder now. I suspect you can now simply wait until this issue is closed, and in the next release the compilation on Sparc will work.
|
|
Also fails in buildbot, looks the same: http://buildbot.askmonty.org/buildbot/builders/sol10-sparc/builds/347/steps/compile/logs/stdio
gmake[2]: Entering directory `/u0/people/bbt/mdb/solaris-sparc/build/mariadb-5.5.47'
|
Undefined first referenced
|
symbol in file
|
rw_pr_rdlock CMakeFiles/strings-t.dir/strings-t.c.o
|
rw_pr_unlock CMakeFiles/strings-t.dir/strings-t.c.o
|
rw_pr_wrlock CMakeFiles/strings-t.dir/strings-t.c.o
|
rw_pr_destroy CMakeFiles/strings-t.dir/strings-t.c.o
|
PSI_server CMakeFiles/strings-t.dir/strings-t.c.o
|
rw_pr_init CMakeFiles/strings-t.dir/strings-t.c.o
|
[ 8%] Building C object mysys/CMakeFiles/mysys.dir/charset-def.c.o
|
[ 8%] Building C object mysys/CMakeFiles/mysys.dir/checksum.c.o
|
[ 8%] Building C object mysys/CMakeFiles/mysys.dir/array.c.o
|
[ 8%] Building C object mysys/CMakeFiles/mysys.dir/charset.c.o
|
strdup_root CMakeFiles/strings-t.dir/strings-t.c.o
|
ld: fatal: symbol referencing errors. No output written to strings-t
|
I've added 5.5 to 'Fix version(s)', because the build is currently attempted on 5.5 in buildbot. We might choose not to fix it for 5.5, but then we need to disable the build.
|
|
On 10.1 it's a bit different:
cc: Warning: -s conflicts with -xdebuginfo=line,param,variable,tagtype,codetag,decl. -s turned off
|
Undefined first referenced
|
symbol in file
|
encryption_handler CMakeFiles/comp_sql.dir/comp_sql.c.o
|
ld: fatal: symbol referencing errors. No output written to comp_sql
|
gmake[2]: *** [scripts/comp_sql] Error 2
|
gmake[2]: Leaving directory `/u0/people/bbt/mdb/solaris-sparc/build/mariadb-10.1.10'
|
gmake[1]: *** [scripts/CMakeFiles/comp_sql.dir/all] Error 2
|
http://buildbot.askmonty.org/buildbot/builders/sol10-sparc/builds/349/steps/compile/logs/stdio
|
|
Ok, this happens because of the following:
- We compile with CFLAGS="... -xO4... "
- That goes into CMAKE_C_FLAGS
- CMAKE_BUILD_TYPE is not set, so it is RelWithDebInfo by default
- CMAKE_C_FLAGS_RELWITHDEBINFO has -xO2 by default.
- On the compiler line CMAKE_C_FLAGS_RELWITHDEBINFO is added after CMAKE_C_FLAGS, so compiler works with -xO2
- cc from sun studio apparently only supports function inlining at -xO4
- we have a check for a working inline keyword in configure.cmake, but these tests only use CMAKE_C_FLAGS
- So the inline test passes (under -xO4), but compilation fails (under -xO2)
As a workaround, one can set cmake -DCMAKE_C_FLAGS_RELWITHDEBINFO="... -xO4..."
Also I'll try to fix the configure.cmake check to detect that inline is broken at configure time, not during compilation
|
|
Having fixed the above, the next error is
"sql/gcalc_slicescan.h", line 191: Warning: Anonymous struct is being declared.
|
"sql/gcalc_slicescan.h", line 202: Warning: Anonymous struct is being declared.
|
"sql/gcalc_slicescan.h", line 208: Warning: Anonymous struct is being declared.
|
"sql/gcalc_slicescan.h", line 209: Warning: Types cannot be declared in anonymous union.
|
"sql/gcalc_slicescan.h", line 212: Error: The name Gcalc_heap::Info::union::___AUD is unusable except in a non-static member function.
|
"sql/gcalc_slicescan.h", line 214: Error: The name Gcalc_heap::Info::union::___AUD is unusable except in a non-static member function.
|
"sql/gcalc_slicescan.h", line 386: Error: The name Gcalc_heap::Info::union::___AUD is unusable except in a non-static member function.
|
3 Error(s) and 4 Warning(s) detected.
|
|
|
trying to compile mariadb-5.5.48 out of the box using
cmake . -L \
|
-DCMAKE_BUILD_TYPE=RelWithDebInfo \
|
-DCMAKE_C_COMPILER=cc \
|
-DCMAKE_CXX_COMPILER=CC \
|
-DCMAKE_C_FLAGS_RELWITHDEBINFO='-g -DNDEBUG -xtarget=ultra2 -xarch=sparcvis -xO4 -xstrconst -mt' \
|
-DCMAKE_CXX_FLAGS_RELWITHDEBINFO='-g -DNDEBUG -xtarget=ultra2 -xarch=sparcvis -xO4 -noex -mt' \
|
-DCMAKE_INSTALL_PREFIX=/opt/fli/mysql \
|
-DINSTALL_SBINDIR=sbin \
|
-DDEFAULT_SYSCONFDIR=/opt/fli/mysql/etc \
|
-DENABLED_LOCAL_INFILE=1 \
|
-DMYSQL_UNIX_ADDR=/tmp/.mysql.sock \
|
-DWITH_EXTRA_CHARSETS=complex \
|
-DWITH_READLINE=ON \
|
-DWITH_MYSQLD_LDFLAGS=-lmtmalloc
|
indeed prevents -xO2 from sneaking in into options that are actually used in compiling, but the linker error
Undefined sumbol first referenced in file
|
rw_pr_rdlock CMakeFiles/strings-t.dir/strings-t.c.o
|
rw_pr_unlock CMakeFiles/strings-t.dir/strings-t.c.o
|
rw_pr_wrlock CMakeFiles/strings-t.dir/strings-t.c.o
|
rw_pr_destroy CMakeFiles/strings-t.dir/strings-t.c.o
|
PSI_server CMakeFiles/strings-t.dir/strings-t.c.o
|
rw_pr_init CMakeFiles/strings-t.dir/strings-t.c.o
|
strdup_root CMakeFiles/strings-t.dir/strings-t.c.o
|
ld: fatal: symbol referencing errors. No output written to strings-t
|
gmake[2]: *** [unittest/strings/strings-t] Error 2
|
is still showing up
|
|
RIght, my comment was incomplete. Also -g disables inlining. But it now works in buldbot. We use this command:
export LDFLAGS=-s
|
cmake . -L -DCMAKE_BUILD_TYPE=Release \
|
-DCMAKE_C_COMPILER=cc \
|
-DCMAKE_CXX_COMPILER=CC \
|
-DCMAKE_C_FLAGS_RELEASE="'-DNDEBUG -xtarget=ultra2 -xarch=sparcvis -xO4 -xstrconst -mt'" \
|
-DCMAKE_CXX_FLAGS_RELEASE="'-DNDEBUG -xtarget=ultra2 -xarch=sparcvis -xO4 -noex -mt'" \
|
-DCMAKE_INSTALL_PREFIX=/opt/fli/mysql \
|
-DINSTALL_SBINDIR=sbin \
|
-DDEFAULT_SYSCONFDIR=/opt/fli/mysql/etc \
|
-DENABLED_LOCAL_INFILE=1 \
|
-DMYSQL_UNIX_ADDR=/tmp/.mysql.sock \
|
-DWITH_EXTRA_CHARSETS=complex \
|
-DWITH_MYSQLD_LDFLAGS=-lmtmalloc
|
Note, I no longer set CC, CXX, CFLAGS and CXXFLAGS in the
environment.
|
|
Hello,
You went faster.
I am going to test it in the next days.
Best Regards,
Pascal
----- Mail original -----
De: "Sergei Golubchik (JIRA)" <jira@mariadb.atlassian.net>
À: pviltard@free.fr
Envoyé: Mardi 16 Février 2016 16:37:00
Objet: [JIRA] (MDEV-8448) cannot build on solaris 10 Sparc
[ https://mariadb.atlassian.net/browse/MDEV-8448?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=80935#comment-80935 ]
Sergei Golubchik commented on MDEV-8448:
----------------------------------------
RIght, my comment was incomplete. Also -g disables inlining. But it now works in buldbot. We use this command:
export LDFLAGS=-s
|
cmake . -L -DCMAKE_BUILD_TYPE=Release \
|
-DCMAKE_C_COMPILER=cc \
|
-DCMAKE_CXX_COMPILER=CC \
|
-DCMAKE_C_FLAGS_RELEASE="'-DNDEBUG -xtarget=ultra2 -xarch=sparcvis -xO4 -xstrconst -mt'" \
|
-DCMAKE_CXX_FLAGS_RELEASE="'-DNDEBUG -xtarget=ultra2 -xarch=sparcvis -xO4 -noex -mt'" \
|
-DCMAKE_INSTALL_PREFIX=/opt/fli/mysql \
|
-DINSTALL_SBINDIR=sbin \
|
-DDEFAULT_SYSCONFDIR=/opt/fli/mysql/etc \
|
-DENABLED_LOCAL_INFILE=1 \
|
-DMYSQL_UNIX_ADDR=/tmp/.mysql.sock \
|
-DWITH_EXTRA_CHARSETS=complex \
|
-DWITH_MYSQLD_LDFLAGS=-lmtmalloc
|
Note, I no longer set CC, CXX, CFLAGS and CXXFLAGS in the
environment.
–
This message was sent by Atlassian JIRA
(v7.1.0-OD-06-005#71002)
|
make error.txt