On 5/4/07, Brandon Van Every <address@hidden> wrote:
On 5/1/07, John Cowan <address@hidden
> wrote:
Brandon Van Every scripsit:
> Ew. Not a case use I remember thinking about, so wouldn't shock me if
> there's something wrong. I'll look into it.
It's actually a pretty common circumstance when you are updating
your Chicken regularly from the darcs head. In my case,
Chicken is in /usr/local, so the EXTANT_CHICKEN winds up being
/usr/local/bin/chicken-static. Then I darcs pull, ccmake, make, and
make install, so the new version of chicken-static winds up overwriting
the old. That's where the Linux version can't cope.
A straightforward fix would be to always copy EXTANT_CHICKEN to /tmp (or
C:\windows\temp) and run it from there.
Why do I have a sinking feeling about that? Is it just that I don't want to write the code for the dynamic case, or some greater can of worms?
I couldn't reproduce the behavior on MinGW when installing over EXTANT_CHICKEN. I think I probably did everything correctly in the 1st place,
i.e. EXTANT_CHICKEN is not a dependency, and this is a Linux specific bug. I will try Cygwin later on to see if it's fine.
I have found that on Cygwin, if I build with /usr/local/bin/chicken-static.exe, with the intent of installing Chicken in /usr/local, chicken-static.exe is never built and never installed. The static libs are built but not installed, which I find even more peculiar. I get the same behavior on MinGW with respect to C:\Program Files\Chicken\bin\chicken-
static.exe. Perhaps CMake 2.4.6 has ad-hoc handling of circular installation dependencies. Linux is implementing a different policy than Cygwin and MinGW. I'll try out CMake CVS soon enough, but I can't make more progress on this until my internet connectivity is restored. I'm using someone else's computer to send this.
Lack of static was an unrelated bug. There's a stray (SET MACOSX TRUE) statement in the 2.6 tarball, which turns off static exes and libs on all platforms, not just MACOSX. I've fixed it in Darcs head.
Cheers,
Brandon Van Every