Hi, I've been comparing testsuite results with PHP compiled in 32 bits vs PHP compiled in 64 bits in Linux on powerpc.
I see some tests which fail in ppc64 but don't fail in ppc32. I'd like to ask your opinion regarding their severity. By severity I mean how likely it is that the failures will affect the average webapp. I have my opinion (see below), but I have to admit I don't have much experience with PHP, so I turn to this list for advice. Here are the failures I'm referring to: Zend/tests/int_overflow_64bit.phpt - integer doesn't overflow: Should go to -9223372036854775808 (most negative 64-bit integer), but stays at 9223372036854775807 (most positive 64-bit integer) ext/gmp/tests/021.phpt - gmp_gcd (greatest common divisor) returns one of it's arguments when the other is zero (only in one case). It should return 0. ext/gmp/tests/026.phpt - gmp_cmp is returning -1 when comparing a very big number to zero. Should return -2. ext/pdo_sqlite/tests/bug_34630.phpt - returns NULL for value associated with :blob ext/pdo_sqlite/tests/bug_39398.phpt - returns NULL for value associated with :boolean ext/pdo_sqlite/tests/bug_39656.phpt - not sure, apparently select query returns nothing? ext/sqlite/tests/pdo/bug_34630.phpt - returns weird string for value associated with :blob ext/standard/tests/array/array_slice_variation6.phpt - returns empty slice in one case In my opinion, the gmp and sqlite failures are not important for the case I have in mind (common webapp). GMP is (AFAIK) geared towards scientific computing, and for sqlite, I believe a webapp would use a server like MySQL or PostgreSQL instead. This leaves me with the integer overflow failure and the array slice failure, but they seem to be very specific, and wouldn't normaly bite the PHP developer. Do you agree? -- []'s Thiago Jung Bauermann IBM Linux Technology Center -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
