I tested 3.6.12 also, and it doesn't build either. (even though it had built previously in the past) I suspect it may have something to do with latest updateto glibc-2.17.0. The samba config does not detect availability of aio suspend64. And that is also where the failure takes place when linking. Builds ok with aio USE flag disabled. --- Linking non-shared library bin/libaddns.a Compiling lib/system.c Linking shared library bin/libnetapi.so.0 lib/system.o: In function `sys_aio_read': system.c:(.text+0x13b1): undefined reference to `aio_read64' lib/system.o: In function `sys_aio_write': system.c:(.text+0x13c1): undefined reference to `aio_write64' lib/system.o: In function `sys_aio_return': system.c:(.text+0x13d1): undefined reference to `aio_return64' lib/system.o: In function `sys_aio_cancel': system.c:(.text+0x13e1): undefined reference to `aio_cancel64' lib/system.o: In function `sys_aio_error': system.c:(.text+0x13f1): undefined reference to `aio_error64' lib/system.o: In function `sys_aio_fsync': system.c:(.text+0x1401): undefined reference to `aio_fsync64' lib/system.o: In function `sys_aio_suspend': system.c:(.text+0x1411): undefined reference to `aio_suspend64' collect2: error: ld returned 1 exit status make: *** [bin/libnetapi.so.0] Error 1 Reproducible: Always
1) Please attach your complete build log. 2) Please reply with the output of `emerge --info =net-fs/samba-3.6.13`.
Created attachment 344830 [details] Build Log
Created attachment 344832 [details] emerge info Please note that the emerge info shows aio USE flag disabled since the last successful build had it manually set that way. Normally it is enabled.
Cannot reproduce on ~amd64, even with glibc 2.17. I wonder why in your build log some parts get built as non-shared (i.e. static) libraries?
This image was built on in a KVM/QMEU environment, and then deployed on a VMware host. Is there anything in the configure stage for samba that a VMware environment could influence somehow?
Think I may have found something. Every time I try to emerge samba, a conftest segfault shows up in dmesg similar to: conftest[13905]: segfault at 1 ip 0000000000000001 sp 00007fff038f9928 error 4 in conftest[400000+1000]
I can see exactly the same error. After commenting out LDFLAGS in my make.conf it compiles fine and I can use the samba client functionality without problems. My LDFLAGS was "-Wl,-O1". Please try without LDFLAGS set.
Nice catch. Linker optimizations would have been the last thing I would have considered. First time I've ever seen this linker flag cause any sort of issue. Confirmed that it works for me.
With the recent package bump in portage, I ran into problems again. I had set up in portage as custom build environment for building samba where is set LDFLAGS="" Well, that fails for me the same as it did earlier. I've found that it's not enough just to redefine LDFLAGS, it must not be set at all. I think this has less to do with linking errors resulting from the linker optimization flag, and probably has more to do with a broken build system method of appending text to LDFLAGS when it has already been defined.
Same here with net-fs/samba-3.6.14-r1 USE="aio client cups fam netapi pam readline server smbclient winbind"
(In reply to comment #10) > Same here with net-fs/samba-3.6.14-r1 USE="aio client cups fam netapi pam > readline server smbclient winbind" BTW, I could emerge with LDFLAGS=""
Created attachment 348604 [details] Prakash's emerge --info
Please reopen if this is still an issue with samba-4.x