Don't know exactly why, but building ncurses on g/fbsd makes it fails as it tries to use system's tic command to build terminfo data. The attached patch fixes the problem, as it makes clear to configure which CC and HOSTCC are used, so that it can really find out that is not crosscompiling. The modified ebuild works fine for me on linux, too. HTH, Diego
Created attachment 57464 [details, diff] ncurses-5.4-r6.patch
i havent really looked but i'm 99% sure that patch is wrong ncurses use $(BUILD_CC) to build executables it will then run ... and HOSTCC is just the old name autotools used for BUILD_CC
Makefiles still uses HOSTCC, that's why I passed it. The patch then can be wrong and needs to be inverted, sorry about that, I have never set up a crosscompiler and I tried to understand how it was working with lu_zero but seems like we misunderstood the order. It should be inverted, so?
no, the makefile sets HOSTCC to BUILD_CC: misc/Makefile.in:HOSTCC = @BUILD_CC@ i dont see why anything needs to be forced other than what is in the ebuild now ...
Mainly because ncurses on g/fbsd finds 'gcc' as $CC instead of i686-*-gcc.
so ?
so CC != HOSTCC and it tries to use the system's tic command which isn't there (in a plain fbsd system, in a 5.3 g/fbsd system or in a non-bootstrapped 5.4 (experimental) g/fbsd system), failing the compilation.
why dont you paste an actual build log so i can investigate further ? :p
Can you wait till tomorrow? My g/fbsd box is off right now as it's night and it does a lot of noise.
Created attachment 61587 [details] ncurses-configure-output Forgot about that, sorry.
i dont see any error in that log ... i do see this though: checking whether we are cross compiling... no
The problem is that one of the makefiles (just don't remember which one right now) refuses to use the just compiled tic because HOST_CC != CC, considering that a kind of crosscompiling.
Well not a problem at this point, as tic is in the base stage, so this can be closed for now.
