Configure tests fail with old version of ld

Configuring ntpsec 0.9.7 (or master) fails on Mac OS X 10.6.8 Snow Leopard with llvm-gcc-4.2:

--- Configuring host --- 
Checking for 'clang' (C compiler)        : not found 
Checking for 'gcc' (C compiler)          : /usr/bin/llvm-gcc-4.2 
Checking compiler version                : 4.2.1 
Checking for program 'bison'             : /opt/local/bin/bison 
Checking for program 'awk'               : /usr/bin/awk 
Checking for program 'sh'                : /bin/sh 
Checking for program 'asciidoc'          : /opt/local/bin/asciidoc 
Checking for asciidoc version >= 8.6.0   : 8.6.9 
Checking for program 'a2x'               : /opt/local/bin/a2x 
Checking for program 'xsltproc'          : /opt/local/bin/xsltproc 
Building version                         : 0.9.7 
--- Configuring main --- 
Checking for endianness                  : little 
Checking if C compiler supports PIC      : no 
Checking if C compiler supports PIE      : no 
Checking if C compiler supports gnu99    : no 
Checking if C compiler supports unused   : no 
Checking if C compiler supports LTO      : no 
Checking if ld compiler supports z_now   : no 
Checking for type uint64_t               : not found 
Checking for type struct if_laddrconf    : not found 
Checking for type struct if_laddrreq     : not found 
Checking for type struct timex           : not found 
Checking for type struct ntptimeval      : not found 
Checking for header stddef.h             : not found 
Checking for field time_tick in struct timex : not found 
Checking for field modes in struct timex     : not found 
Checking for field time.tv_nsec in struct ntptimeval : not found 
Checking for field tai in struct ntptimeval          : not found 
Checking sizeof long                                 : The configuration failed

Looking in the config.log, all the configure tests are failing for the same reason, which looks like this:

err: ld: unknown option: -o/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_sysutils_ntpsec/ntpsec/work/ntpsec-0.9.7/build/.conf_check_3137299a714cbee496d5031eb7b85c5c/testbuild/main/testprog
collect2: ld returned 1 exit status

I guess old ld wants there to be a space between -o and the path. I don't know how to change this in your configuration system.

@ryandesign can you attach build/config.log?

main.log

config.log

I put a space in there and then it fails for me:

['/usr/bin/gcc', 'test.c.1.o', '-o /u1/src/NTP/ntpsec/build/.conf_check_bf498218
6e9f029b345b3c725419d2f7/testbuild/main/testprog', '-Wl,-Bstatic', '-Wl,-Bdynami
c', '-lm']
err: /usr/lib/gcc/x86_64-pc-linux-gnu/6.3.0/../../../../x86_64-pc-linux-gnu/bin/ldonf_check_bf4982186e9f029b345b3c725419d2f7/testbuild/main/testprog: No such file or directory
collect2: error: ld returned 1 exit status

The -o comes from this line in .waf-1.9.7-fc95bab69065dbc43ddd71d4b69d0078/waflib/Tools/gcc.py

    v.CCLNK_TGT_F=['-o']

I can build fine on up to date macOS:

Apple LLVM version 8.0.0 (clang-800.0.42.1)

mentioned in commit 977517bc

Alas. I can see no better way around this than documenting how and when to tweak the waf installation in INSTALL.

closed

Thanks, but that doesn't help. waf will still create only a single command line argument, consisting of dash-o-space-path, which doesn't work with any compiler. I guess waf needs to be modified to create two command line arguments, one just dash-o, the other just the path. I guess this is a bug in waf:

https://github.com/waf-project/waf/pull/2003

Once they release a version of waf with the fix, you'll have to update your bundled copy of waf.

reopened

I read the waf bug, does not sound like they will be pulling a patch soon. Is there anything we could do to work around this?

Is the problem on OS X? Could we detect OS X and then tell waf how to output -o differently?

@ryandesign So the resolution is that this is a waf bug which (a) we can't fix, and which (b) is in process upstream.

Therefore (and because I'm trying to clear our issue list of clutter so we can focus on what we can fix), I'm going to close this again. When you wrestle ita into accepting a fix, ping us and we'll update our waf.

closed