Waldek Hebisch <[EMAIL PROTECTED]> writes: > But I must admit that I saw quite a few weird build problems -- most of the > time problem went away re-running make but some were repeatable.
Oh dear. So, why doesn't this happen with "GoldenAxiom"? Is the new build procedure more sensitive to sloppy programming bugs? > I was able to track some problems to real bugs (at some point Axiom generated > wrong data structure, causing memory corruption later and finally a weird > crash), but some went away due to seemingly unrelated changes (and most is > not repeatable). Since I suspect that we still may corrupt memory I tried to > build Axiom setting GCL safety to 3. After 16 hours the build stopped in > guessing package with report about symbol having undefined value (I still > have to investigate this deeper), but the previous parts build OK. Well, meanwhile I was able to build most of axiom by building and installing GCL (from build-improvements) separately. Before that, I almost consistently got the error about missing ALIST.o. Might that help to track down the bug? However, now the build stops in guess-pkg: FAMR2.NRLIB/code.o is missing. Looking at trace, the problem seens to start in SUPEXPR, I get a "non extending category" message there, albeit without any error message. The next domain which is compiled is FAMR2, and there it says System error: Cannot create the file FAMR2.erlib/index.KAF. (Note that FAMR2 does *not* depend on SUPEXPR...) Martin _______________________________________________ Axiom-developer mailing list [email protected] http://lists.nongnu.org/mailman/listinfo/axiom-developer
