campbell@Thalatta.COM (Bill Campbell) (08/08/90)
Sender: Reply-To: campbell@Thalatta.COM (Bill Campbell) Followup-To: Distribution: world Organization: Celestial Software, Mercer Island, WA Keywords: perl, Xenix, gcc I am new to this group and to perl. I saw a posting from someone late last week describing problems with SCO Open Desktop where his make test failed on op.split. I have the same problem, but was waiting for the answer to be posted. Anyway op.split fails after test # 4 completes successfully, aborting with a core dump. I am running Xenix 2.3.3 with the lng085 update and compiling with gcc. (I am at patchlevel 18). -- ....microsoft--\ Bill Campbell; Celestial Software ...uw-entropy----!thebes!camco!bill 6641 East Mercer Way ....fluke------/ Mercer Island, Wa 98040 ....hplsla----/ (206) 232-4164
6sigma2@polari.UUCP (Brian Matthews) (08/08/90)
In article <5332@thebes.Thalatta.COM> campbell@Thalatta.COM (Bill Campbell) writes: |Anyway op.split fails after test # 4 completes successfully, |aborting with a core dump. I've got perl running on SCO Open DeskTop, but not until solving the op.split and other problems. The first was a yacc problem. Yacc has to be passed a -Sm10000 flag. The README mentions this for Xenix, but it's also true for Open DeskTop (and undoubtedly SCO's SVR3.2). When Configure asks you for yacc or bison, just type in "yacc -Sm10000" (or be smart and use bison). Larry: the README should probably mention SCO's SVR3.2 and ODT along with Xenix. The second problem is the op.split problem. There's code in the source file that does splits (split.c?) that does one thing on 286s and another thing on real machines. The problem is that cc defines M_I86, even though you're on a 386, which split.c uses (actually M_I86 is used to set another variable which split.c uses) to do the 286 code, which blows up. The solution is to enter -UM_I86 when Configure asks you for any additional cc flags. The third problem I encountered was odd out of memory problems and core dumps within malloc. I reConfigured perl to use the malloc that comes with perl and all is fine. Note that none of the regression tests spotted this. Hope this helps. -- Brian L. Matthews blm@6sceng.UUCP