Bug #39080 | Compile errors in Cluster: LockQueue.hpp + PosixAsyncFile.cpp + more | ||
---|---|---|---|
Submitted: | 27 Aug 2008 19:15 | Modified: | 6 May 2009 9:04 |
Reporter: | Joerg Bruehe | Email Updates: | |
Status: | Closed | Impact on me: | |
Category: | MySQL Cluster: Cluster (NDB) storage engine | Severity: | S1 (Critical) |
Version: | mysql-5.1-telco-6.3 -> 6.3.17 Cluster | OS: | Other (See text) |
Assigned to: | jack andrews | CPU Architecture: | Any |
[27 Aug 2008 19:15]
Joerg Bruehe
[28 Aug 2008 7:09]
Joerg Bruehe
I have run "make -k" twice on the affected hosts, the logs are saved. Log file names etc. are added in a private entry.
[1 Sep 2008 18:40]
Joerg Bruehe
I applied the patch provided from development: - Solaris 9 now built on Sparc-64bit and on 386, - but it still failed on Sparc-32bit; - HP-UX failed again. In detail: ==== HP-UX 11.23 ==== "ndbfs/PosixAsyncFile.cpp", line 458: error #2101: "seek_val" has already been declared in the current scope off_t seek_val; ^ 1 error detected in the compilation of "ndbfs/PosixAsyncFile.cpp". Checking the source, it is obvious there is an error: 441 int PosixAsyncFile::readBuffer(Request *req, char *buf, 442 size_t size, off_t offset) 443 { 444 int return_value; 445 req->par.readWrite.pages[0].size = 0; 446 #if ! defined(HAVE_PREAD) 447 off_t seek_val; 448 if(!use_gz) 449 { 450 while((seek_val= lseek(theFd, offset, SEEK_SET)) == (off_t)-1 451 && errno == EINTR); 452 if(seek_val == (off_t)-1) 453 { 454 return errno; 455 } 456 } 457 #endif 458 off_t seek_val; 459 if(use_gz) 460 { 461 while((seek_val= azseek(&azf, offset, SEEK_SET)) == (off_t)-1 462 && errno == EINTR); 463 if(seek_val == (off_t)-1) 464 { 465 return errno; 466 } 467 } 468 I suspect lines 458 .. 467 should be the "#else" block of the "#if" in line 446, but didn't yet try that. ==== Solaris 9, Sparc (32 bit) ==== The linker fails on an input library without reporting any details, to me it seems to be an internal error: CC -g -DSAFE_MUTEX -g -noex -mt -D_FORTEC_ -xarch=v8 -D_FILE_OFFSET_BITS=64 -DHAVE_RWLOCK_T -xarch=v8 -o gen_lex_hash gen_lex_hash.o ../vio/libvio.a ../mysys/.libs/libmysyswrap.a -lpthread -lthread -lpthread -lthread -lpthread -lthread -lpthread -lthread ../mysys/.libs/libmysyslt.a -lpthread -lthread -lpthread -lthread -lpthread -lthread -lpthread -lthread ../dbug/.libs/libdbuglt.a -lpthread -lthread -lpthread -lthread -lpthread -lthread -lpthread -lthread ../regex/libregex.a ../strings/.libs/libmystringsextra.a -lpthread -lthread -lpthread -lthread -lpthread -lthread -lpthread -lthread ../strings/.libs/libmystringslt.a -lpthread -lthread -lpthread -lthread -lpthread -lthread -lpthread -lthread ../zlib/.libs/libzlt.a -lrt -lpthread -lthread -lposix4 -lresolv -lgen -lsocket -lmtmalloc -lnsl -lm -lpthread -lthread ld: elf error: file ../mysys/.libs/libmysyswrap.a: elf_getarsym ld: fatal: File processing errors. No output written to gen_lex_hash gmake[2]: *** [gen_lex_hash] Error 1 I cannot spot any significant difference to the platforms where it works.
[9 Jan 2009 18:47]
Magnus BlÄudd
Most likely the "off_t seek_val" variable should simply be moved to begininig of function.
[15 Apr 2009 10:24]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/72119 2937 jack andrews 2009-04-15 bug#39080 Compile errors in Cluster: LockQueue.hpp + PosixAsyncFile.cpp + more
[16 Apr 2009 16:53]
Jon Stephens
Hi Jack, I can't document a bugfix without knowing which versions the fix went into. The bug report should be updated with this info + set into Documenting status automatically when your pushes are merged to production trees. If this isn't happening for your pushes, please check with one of the other devs (I suggest Jonas or Magnus) to get help in figuring out why. Thanks!
[29 Apr 2009 10:13]
jack andrews
this is fixed in 6.3 and 7.0
[29 Apr 2009 11:37]
Bugs System
Pushed into 5.1.32-ndb-6.3.25 (revid:jack@sun.com-20090429081724-ny49e0kw3y6xpjkw) (version source revid:jack@sun.com-20090415102342-n4p58kxi1skfhqgd) (merge vers: 5.1.32-ndb-6.3.25) (pib:6)
[30 Apr 2009 8:51]
Bugs System
Pushed into 5.1.32-ndb-7.0.6 (revid:ole.john.aske@sun.com-20090430084504-2c8iheeuc631njs5) (version source revid:ole.john.aske@sun.com-20090430084504-2c8iheeuc631njs5) (merge vers: 5.1.32-ndb-7.0.6) (pib:6)
[6 May 2009 9:04]
Jon Stephens
Documented bugfix in the NDB-6.3.25 and 7.0.6 changelogs as follows: NDBCLUSTER did not build correctly on Solaris 9 platforms. Closed.