Aleksandar Kuktin wrote: >> On Sat, 30 Mar 2013 16:23:43 -0500 >> Bruce Dubbs <bruce.du...@gmail.com> wrote: >> >> I've finally figured out what is making the mangle33.C program fail. >> It is an arbitrary limitation in tcl. This particular gcc test >> creates a c++ namespace *name* of 4044 characters in length. It >> turns out that there is a value buried deep in tcl code >> (generic/regc_nfa.c): >> >> /* >> * Arbitrary depth limit. Needs tuning, but this value is sufficient >> to >> * make all normal tests (not reg-33.14) pass. >> */ >> #define DUPTRAVERSE_MAX_DEPTH 500 >> >> Changing this to 5000 makes the gcc test pass. Is it worthwhile to >> make this change in the Chapter 5 build of tcl? How about BLFS?
> It is worthwhile to hack tcl to not use an arbitrary value, as per the > zero-one-infinity rule. I think tcl is trying to prevent infinite recursion. Actually I'm not convinced a 4044 regex pattern is reasonable, even if it doesn't have any meta characters. For this test, a regex is not required, but a simple strstr. Trying to put that into the gcc test harness is actually quite a bit more difficult than changing tcl. Changing tcl to be able to accept a configuration value for this might be possible, but a lot of work for what looks like only minimal value. > However, as that is a lot of work and also something the developer > should have done, changing DUPTRAVERSE_MAX_DEPTH to 5000 (or whatever) > is, IMHO, worthwhile. Yes, I'm leaning that way. I'd like to figure out he libmudflap failures too. -- Bruce -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
