Hi, Solène RAPENNE <[email protected]> writes:
> Jeremie Courreges-Anglas writes: > >> Solène Rapenne <[email protected]> writes: > >>> I am using a commit of github because the latest release on github >>> doesn't compile so I need to take the latest commit archive. Is it >>> possible to write something cleaner than setting the DISTFILES with the >>> commit name ? >> >> I'd suggest using something like: >> >> DISTNAME = bastet-0.43.1.2015083000 >> >> and getting rid of DISTFILES = ... >> > > I found a cleaner solution, I posted a mail two minutes ago with a new > version of the Makefile, I didn't see your mail before posting mine. > I use GH_COMMIT to have the commit, then when downloading, it sets the > name of the archive downloaded to the distname with .tar.gz as > suffix. So I don't need to declare DISTFILES and the DISTNAME is clean. The 2015083000 part in 0.43.1.2015083000 was to make clear that it's the 0.43.1 release plus a few patches. I don't think we actually care. >>> It compiles ONLY with clang, for unknown reason. >>> >>> I used portcheck -N and I fixed the complaints. >> >> Looks fine to me, but... it looks like a patch is missing: >> >> ===> Building for bastet-0.43.1.2015083000 >> c++ -O2 -pipe -DNDEBUG -Wall -I"/usr/local"/include -DLOCALSTATEDIR="/var" >> -c Ui.cpp >> Ui.cpp:479:2: error: use of undeclared identifier 'usleep' >> usleep(500000/6); >> ^ >> 1 error generated. >> *** Error 1 in >> /usr/obj/pobj/bastet-0.43.1.2015083000/bastet-5037e410fde6fafe8e6bc4cf17a12ad98e01c17e >> (<sys.mk>:124 'Ui.o') >> *** Error 1 in . (/usr/ports/infrastructure/mk/bsd.port.mk:2676 >> '/usr/obj/pobj/bastet-0.43.1.2015083000/.build_done') >> *** Error 1 in /usr/ports/mystuff/games/bastet >> (/usr/ports/infrastructure/mk/bsd.port.mk:2398 'all') >> >> ... using llvm-3.8.0 on -current. Also, >> > > This is weird, on 5.9 there is no such issue. Maybe it's an issue of > llvm in current ? > >>> MAKE_FLAGS = CPPFLAGS="-I\"${PREFIX}\"/include \ >>> -DLOCALSTATEDIR=\"${LOCALSTATEDIR}\"" \ >>> LDFLAGS="-L\"${PREFIX}\"/lib -lcurses -lboost_program_options" >> >> while it's not really incorrect, we tend to use LOCALBASE for "where >> other ports install stuff" and PREFIX for "where this port installs >> stuff". While I don't really understand the purpose of having two >> variables, it's better to stay consistent. So that would be LOCALBASE >> here. >> > > In the latest version of the port I have been patching bastet's Makefile > to add the CPPFLAGS and LDFLAGS and I used LOCALBASE in it instead of PREFIX. Patching is fine with tangled Makefiles, but we tend to avoid patches when MAKE_FLAGS is enough. >>> NO_TEST = Yes >>> >>> do-configure: >>> ${SUBST_CMD} ${WRKSRC}/Config.cpp >> >> This doesn't do much now. IIRC I wanted to make sure that we respected >> ${LOCALSTATEDIR}, which is supposed to be a user setting. I doubt that >> it is actually supported in practice... > > I will remove the SUBT_CMD on Config.cpp then. > >> >> Updated tarball attached. Looks like the Enter key doesn't work as >> advertised? > > The game works fine on 5.9 amd64. Is the Enter key inactive in the > menu or while playing ? While playing. Could you send an updated tarball? (Not just the Makefile, please.) -- jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
