[Ecls-list] Re: Small fixes and status of CVS

Prut Flut prutflut at m-net.arbornet.org
Mon Jan 12 02:08:02 UTC 2004


I hope that my mail will show up in the correct thread.

Anyway, I was trying this morning to build the current version
of ecls for alpha-unknown-netbsd1.6.1, which indeed failed with
a Segmentation fault - core dumped.  I have not solved completely
the problem, but now I arrive further in the build process: the
problem (looks to me, but I know nothing about 64bit arch) is
that MOST_POSITIVE_FIXNUM is promoted to double in the following
line of code and ends up as -1.0:
[src/c/hash.d: line 447 of 664 [67%]]
        if (delta < 1 || delta > MOST_POSITIVE_FIXNUM) {

This test (delta is 1.5, this is the first call to
cl__make_hash_table, in cl_boot) sends ecls into an infinite
loop calling FEerror and then piling up 16374 stack frames.

When the above line is changed into
[src/c/hash.d: line 447 of 664 [67%]]
        if (delta < 1 || delta > ((unsigned) MOST_POSITIVE_FIXNUM)) {

ecl_min goes on (but only a bit ;-() and stops with
Unrecoverable error: 
Lisp initialization error.

Abort trap - core dumped
gmake[1]: *** [ecl] Error 134
gmake[1]: Leaving directory `/www/p/pcrama/ecls-track/ecl-0.9c/build'
gmake: *** [all] Error 2

Well, I didn't get very far, but it was fun to try!
And I will keep trying when I have a bit more time.

Philippe Crama




More information about the ecl-devel mailing list