On 01/26/2012 11:17 PM, Norbert Thiebaud wrote:
as for the dmake side of thing (solenv/inc/sol*4.mk) something like .IF "$(COM)$(OS)$(CPU)" == "GCCSOLARISX" (for intel x86_64 ) or GCCSOLARISI for x86, etc... .INCLUDE : uxnsolg[x|i|..].mk and create such a file based on similar unxix build... look into bsd and linux ones for template... or even the existing solaris one.. and scrub out stuff that are SunStudio related...
But when you look at the start of solenv/inc/unx.mk, specific unx*.mk for Solaris/GCC are already there (unxsogs.mk for Solaris/GCC/SPARC and unxsogi.mk for Solaris/GCC/x86) and already should be called. To me, Gary's build log looks like the wrong unxsoli4.mk (Solaris/SunStudio/x86) gets used.
On libreoffice-3-5 branch, all the relevant variables used in the conditionals in solenv/inc/unx.mk to branch to a specific unx*.mk (i.e., the variables OS,CPU, COMEX, COM) should be defined in Env.Host.sh. I suspect that COMEX has an odd value.
And indeed, in configure.in it looks like COMEX is nowadays only set on Windows, left empty on all other platforms (that was probably different in the past, and somebody cleaned up too aggressively). So, all the Solaris checks in solenv/inc/unx.mk should probably be rewritten to use COM instead of COMEX. For GCC, COM=GCC (see set_soenv.in), and for Sun Studio COM probably used to be something like CC or CC5, but apparently has also already been scrubbed from set_soenv.in. So that you might just as well also remove the branches for unxsol?4.mk from solenv/inc/unx.mk.
Stephan