Filing a bug based on #Ada@Freenode discussion: Polyorb supports multiple personalities (CORBA, DSA, MOMA). It would be nice if user could configure them (via USE flags for example). Some notes: 1) IIRC, by default only the CORBA personality is compiled. If you need others, you need to supply option --with-appli-perso="corba dsa moma" to the configure script. 2) Some personalities (like DSA) require a specific GNAT version (GNAT GPL 2006 for example). Reproducible: Didn't try
Well, the trial with --with-appli-perso="corba dsa" failed with: Making all in dsa /bin/sh: line 11: cd: dsa: No such file or directory This does not look like something that could be helped by a specific gnat version, it looks like missing code (unless one is supposed to download this part separately. The dir dsa under src is indeed absent). May be it has been added since I did that cvs checkout? Will need to check again (but it is mentioned in the INSTALL file nonetheless). ATM I am testing moma personality, will report how that one worked. I have a few questions though: 1. Is there any problem with just building all personalities? I can make them depend on some local USE flags, but if the difference is only in saving a few MB I am not sure it is really worth it. 2. There is also --with-proto-perso= flag. These things are also called personalities, although without the qualifier "application". What are these? Or what is the difference from application personalities? Similarly, should they all be just enabled or should they be made configurable? George
(In reply to comment #1) > /bin/sh: line 11: cd: dsa: No such file or directory [snip] > part separately. The dir dsa under src is indeed absent). May be it has been > added since I did that cvs checkout? You should have src/dsa if you have checked out the latest source from CVS. If you look at https://libre.adacore.com/cvsweb/polyorb/src/dsa/ you see that some files there are 2 years old. > 1. Is there any problem with just building all personalities? Probably it is fine if you always build all personalities. The compile time might be longer, but I haven't measured how much longer. > 2. There is also --with-proto-perso= flag. These things are also called > personalities, although without the qualifier "application". What are these? --with-proto-perso defines used communication protocols, like GIOP or SOAP. Again, it is probably ok to compile them all.
Strange, I am trying to build polyorb with additional protocols and personalities and, some 40 min into the build, I am hitting this: gnatbind -aO./ -I. -I../../src -I../../src -I../../src/setup -I../../src/setup -I../../src/giop -I../../src/giop -I../../src/giop/iiop -I../../src/giop/iiop -I../../src/giop/diop -I../../src/giop/diop -I../../src/giop/iiop -I../../src/giop/miop -I../../src/giop/miop -I../../src/giop/iiop/ssliop -I../../src/giop/iiop/ssliop -I../../src/soap -I../../src/soap -I../../src/srp -I../../src/srp -I../../src/corba -I../../src/corba -I../../cos/naming -I../../cos/naming -I../../idls/cos/naming -I../../cos/time -I../../cos/time -I../../idls/cos/time -I../../src/moma -I../../src/moma -I../../src/web_common -I../../src/web_common -I../../src/ssl -I../../src/ssl -aI/usr/include/ada/xmlada -aO/usr/lib/ada/x86_64-pc-linux-gnu-gnat-gcc-4.1/xmlada/lib -I- -x po_catref.ali error: "output.adb" must be compiled error: ("/usr/lib/gnat-gcc/x86_64-pc-linux-gnu/4.1/adalib/output.ali" is obsolete and read-only) gnatmake: *** bind failed. make[2]: *** [po_catref] Error 4 make[2]: Leaving directory "/var/tmp/portage/dev-ada/polyorb-2.1.0-r1/work/LocalSource/tools/po_catref" which I don't understand. First I thought this is asis related (as this file is installed by asis), so I rebuilt it (hit a few problems on the way that I had to fix - the gnat-gcc's gnatmake behaves differently from gnat-gpl's for some reason (even though the gnatmake.adb files are equivalent, I even run a diff :), but I did not try to trace it any deeper) regarding how they [re-]compile stuff that has already some visible version installed), but that did not change anything. Now, this message is supposed to be triggered when the source changes, it is spit in bcheck.adb which compares some timestumps. However this should not be the problem - portage timestumps everything upon install and other ali's did not trigger this either. To be absolutely sure I touched that ali file (thus making its timestump definitely a later one, but still got the same message (upon rerunning make in the dir)). Any idea what may be causing this? I guess I should check out polyorb again and try with more recent sources indeed (I wanted to first check how personality moma would build, since the source I had seemed to be fine with just the defaults).. George
A short update. I tried building a fresh copy of polyorb - checked out yesterday, however this seems quite problematic. First, it comes without the ready configure and some Makefile.in files. automake and aclocal were complaining lice crazy, I even had to run libtoolize once to transform some .m4 file. In the end I settled for copying new sources over old ones and trying to reuse the "old build scripts", that is old configure.ac and Makefile.am's. This still fails with complaints that some needed file is missind for creating Makefil.in in src/security. Trying to ignore this error and forcing make breaks right away with apparent whitespace problems in some Makefile trying to call some python script in compilers/idlac. The short summary - apparently the source tree is very unstable. Possibly later checkouts might start to build again (and I had recently a conversation with somebody in #ada on freenode irc who complained that he could not build polyorb from cvs either). Looks like the snapshot I have in has the most stable polyorb version atm. It at least builds :), even thought just the "core" part. I guess this request may be revisited later, but for now it does not seem possible to do much on it. George
(In reply to comment #4) > Trying to ignore this error and forcing make breaks right away with apparent > whitespace problems in some Makefile trying to call some python script in > compilers/idlac. Actually not whitespace, some really odd stuff: make[3]: Entering "/var/tmp/portage/dev-ada/polyorb-2.1.0-r1/work/LocalSource/compilers/idlac" (cd . && \ python make_nodes.py nodes.txt > idl_fe-tree.ada && \ -w idl_fe-tree.ada && \ rm idl_fe-tree.ada) (cd . && \ python make_nodes.py nodes.txt > idl_fe-tree.ada && \ -w idl_fe-tree.ada && \ rm idl_fe-tree.ada) /bin/sh: line 2: -w: command not found /bin/sh: line 2: -w: command not found But no matter. The package as a whole seems to be too far away from anything "release-like" atm. George
removed from the tree
