On Sat, 7 Apr 2012, Nick Couchman wrote: > I'm trying to compile 0.44.1 on CentOS 5, and am running into the following compile-time error: > > g++ -DHAVE_CONFIG_H -I. -I. -I. -I/usr/include/nss3 -I/usr/include/nspr4 -Wall -D__CEPH__ -D_FILE_OFFSET_BITS=64 -D_REENTRANT -D_THREAD_SAFE -D__STDC_FORMAT_MACROS -D_GNU_SOURCE -rdynamic -Winit-self -Wpointer-arith -fno-strict-aliasing -DCEPH_LIBDIR=\"/usr/lib64\" -Wnon-virtual-dtor -Wno-invalid-offsetof -Wstrict-null-sentinel -I../src/leveldb/include -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64 -mtune=generic -MT libos_la-FileStore.lo -MD -MP -MF .deps/libos_la-FileStore.Tpo -c os/FileStore.cc -fPIC -DPIC -o .libs/libos_la-FileStore.o > os/FileStore.cc: In member function 'int FileStore::_zero(coll_t, const hobject_t&, uint64_t, size_t)': > os/FileStore.cc:3013: error: 'fallocate' was not declared in this scope > os/FileStore.cc: In member function 'void FileStore::sync_entry()': > os/FileStore.cc:3385: warning: passing 'float' for argument 1 to 'SyncEntryTimeout::SyncEntryTimeout(int)' > make[3]: *** [libos_la-FileStore.lo] Error 1 > make[3]: *** Waiting for unfinished jobs.... > > Interestingly, if I go to that block of code, there's an "#ifdef HAVE_FALLOCATE" block there, and this is inside that block. So, I go and look at the output of configure and I see: > > checking for utime.h... yes > checking for sync_file_range... no > checking for fallocate... no > checking for a Python interpreter with version >= 2.4... python > checking for python... /usr/bin/python > > Configure script believes fallocate is not available. And, indeed, from config.log: > > configure:18309: checking for fallocate > configure:18309: gcc -o conftest -g -O2 conftest.c >&5 > /tmp/ccybZzwg.o: In function `main': > /root/ceph-0.44.1/conftest.c:81: undefined referenece to `fallocate' > > Okay, so compiler cannot find it, then why is this block of code being evaluated?? I poked around a bit more, and found that HAVE_FALLOCATE is defined in /usr/include/linux/fs.h, which appears to be included in FileStore.cc. So, even though the configure script fails to detect fallocate correctly, it's still pulling it in because it is defined in the linux/fs.h file, even though it doesn't work. What's the best way to fix this build issue and have it correctly disable HAVE_FALLOCATE if fallocate() doesn't really work? That's annoying. You could modify configure.ac so that it defines something other than HAVE_FALLOCATE (CEPH_HAVE_FALLOCATE, perhaps) and adjust the #ifdefs accordingly... sage -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
