Front page | perl.perl5.porters |
Postings from July 2000
Can some of you on Unix systems grep your /usr/include directories and see if you have QUAD_MAX defined, and if so, whether it's a floating point number or not? I'm cross-compiling Perl 5.6.0 in 64-bit mode on HP-UX for PA-RISC and for IA64. On PA-RISC, IV_MAX gets expanded through means of the PERL_QUAD_MAX macro via the QUAD_MAX macro which is not defined on the PA-RISC. This means perl.h defines PERL_QUAD_MAX to be ((IV) (PERL_UQUAD_MAX >> 1)), where PERL_UQUAD_MAX is (~(UV)0). On the IA64 platform, QUAD_MAX is defined in /usr/include/float.h to be LDBL_MAX, which is a nasty 1.xxxE+4932L, and so perl.h defines PERL_QUAD_MAX to be equal to QUAD_MAX and therefore IV_MAX becomes a very large *negative* number. The base/rs.t test thus fails because the DIE in pp_enteriter is called. I need more information about what QUAD_MAX is supposed to be defined as before I can try to fix this. On both PA-RISC and IA64, /usr/include/inttypes.h has INT64_MIN and INT64_MAX macros available. JeffThread Next