On 09/09/15 15:08, Lennart Sorensen wrote: > On Wed, Sep 09, 2015 at 08:27:44AM +0100, Jo Shields wrote: >> It's not the silicon - it failed on a G5 XServe as well >> >> It's not gcc - I did a successful build on gcc5. >> >> It's not the kernel version - it succeeded on a Mac Mini with >> 3.16.0-4-powerpc... actually... You've been using a ppc64 kernel on that >> POWER7, right? > Yes I have. > >> porpora (build daemon 2), powerpc-unicamp-01 (build daemon 1) and partch >> (porterbox) all use 3.16.0-4-powerpc64. i've been having successful >> builds on a G4 Mac Mini with 3.16.0-4-powerpc >> >> I don't have a definite 3.16 64-bit kernel success, or any 32-bit kernel >> failures, in my data set. > So it has failed on power8 machines, and on a 970 (the G5). What CPU > was in the mac mini? > > Hmm. > > CPU |kernel|flavour|libc |monolite version|status|name > ------+------+-------+--------+----------------+------+------- > power8|3.16 |ppc64 |2.19-19 |4.0.4.0 |failed|powerpc-unicamp-01 > power7|3.2 |ppc64 |2.19-19 |4.1.0.0 |works |rceng03 > power7|3.16 |ppc64 |? |4.0.4.0 |failed|partch > mac G5|3.16 |? |? |? |failed|porpora > mini |3.16 |ppc |? |? |works |
porpora has the same ppc64 kernel as partch, libc 2.19-19 > Could it be the monolite version in use? > > My working build says: > Bootstrap compiler: Mono C# compiler version 4.1.0.0 > > The failed log on powerpc-unicamp-01 and partch both say: > Bootstrap compiler: Mono C# compiler version 4.0.4.0 That's an interesting idea. Try building from http://http.debian.net/debian/pool/main/m/mono/mono_4.0.4.1+dfsg-1.dsc which bundles a matched monolite generated from its own sources (get-monolite-latest downloads the most recent available monolite with a compatible ABI) It *could* be related to endian issues in the class library of monolite itself
