You could try compiling with -E or -P to make sure that the #include file is actually being included. If there are #ifdef's around the include, it is possible for it to not be included (and this is done silently). I doubt that is the case here, but it might be worth a try...
max James C. McPherson wrote: > Steve Gonczi wrote: >> Looks like there is a missing extern in >> usr/src/lib/sun_sas/common/devtree_device_disco.c >> >> extern scsi_lun_t scsi_lun64_to_lun(scsi_lun64_t lun64); >> >> The extern above appears in usr/src/lib/libdevid/libdevid.h, (it >> looks like it is #included as <libdevid.h>, but I see no evidence >> of libdevid.h actually being included) . > > libdevid.h is definitely included there, on line 33 > http://src.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/lib/sun_sas/common/devtree_device_disco.c#33 > > > > http://src.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/lib/libdevid/libdevid.h > > > has the extern definition of scsi_lun64_to_lun() on line 65, and the > function itself is defined at > http://src.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/common/devid/devid_scsi.c#1444. > > > > > >> BTW can someone cast some light how an #include statement can >> silently fail? I have seen this occasionally, and include file is >> not found, but there is no error message, I just find out indirectly, >> when things that should be defined go missing. > > Strange. The #include should work, or not work, but never fail silently. > > >> Also, I see what appears to be a build race when doing a full >> nightly. Seems that at the time the build is trying to link >> /opt/SUNWspro/bin/cc -o libsun_sas.so.1 ... >> The libdevid library is not yet built, so the link step is unable to >> resolve the above symbol. >> >> If I do a single threaded build of the libraries using plain make, it >> all works, (given that I add the extern) this is only an issue with a >> multi-threaded build. I am building on an 8 cpu nehalem box. > > How high is your dmake maxjobs setting? How are you running the build? > > We haven't seen this problem on the gate machines, so more information > would be handy to have. > > > James C. McPherson > -- > Senior Kernel Software Engineer, Solaris > Sun Microsystems > http://blogs.sun.com/jmcp http://www.jmcp.homeunix.com/blog > _______________________________________________ > zfs-code mailing list > zfs-code at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-code >