Hi,
On Sat, 27 Nov 1999, Claudio Matsuoka wrote:
> > > In native mode, the configure script
> > > fails to set the appropriate -I directories and headers are not
> > > found.
> > Which of them?
>
> All the relative paths to the package's local include directories. Here
> are the actual parameters it uses:
>
> gcc -g -O2 -I/usr/local/lib/glib/include -I/usr/local/include -c
> savegame.c
> savegame.c:33: engine.h: No such file or directory
> savegame.c:35: heap.h: No such file or directory
> savegame.c:38: graphics_png.h: No such file or directory
> *** Error code 1
Did you use recent versions of those tools? Anyway, adding a
-I@whatever-the-base-path-name-was@/src/include to all Makefile.ams should
fix that; I'll add it asap.
> > Be warned, though- the AC/AM stuff wasn't written with
> > cross-compiling in mind.
> AC seems to be cross-compiling aware, but AM is not that smart. For
> instance, it tries to locate libpng testing for $lib_dir/libpng.so
> instead of using AC_TRY_COMPILE or whatever ($lib_dir is specified by
> --with-png-dir).
That's my fault. I wrote that part of the aclocal.m4 package (and I'm
not exactly an AM wizard). AC_TRY_LINK does sound like a good candidate to
check for that, though; I'll add it.
> NetBSD's PNG lib is /usr/pkg/lib/libpng.so.$version,
> requiring --with-png-dir=/usr/pkg and a link from libpng.so to be
> created manually.
That's not acceptable, as long as there are alternatives.
> > BTW, did you test the build afterwards?
> I didn't complete the build -- I was trying to figure out why the
> autoconfiguration was not working. Does it work out of the box in
> FreeBSD systems?
Not quite; it required some AM tweaking as well. I think that I added all
changes I did to the CVS repository, but I didn't test it afterwards.
However, if it breaks, Rink is sure to complain :-)
> > > And it seems that something is wrong with the timestamps of the
> > > files checked out from the repository.
> > Strange. Which files were affected?
>
> Apparently the files recently checked in were set to "more than one hour
> in the future" (can ls(1) be set to show the time instead of year in
> such situation?). It seems to be a time zone problem (I'm in GMT-2).
I'm in GMT+1, so I'm later (I suppose).
> Shouldn't CVS take care of time zone conversion?
I think so; it might have some problems if the client machine's rtc is
broken, though. My box at home isn't off by more than a few seconds, but
last week I checked in some files from a box at work that is both in the
wrong time zone and half a day off.
Are those problems persistant, or was it just a unique anomaly?
llap,
Christoph
