On 10. Aug 2021, at 14.17, Michael Str?der <michael at stroeder.com> wrote:> > On 8/10/21 11:16 AM, Timo Sirainen wrote: >> On 10. Aug 2021, at 10.33, Michael Str?der <michael at stroeder.com> wrote: >>> >>> On 8/10/21 10:02 AM, Milan P. Stani? wrote: >>>> I'm trying to upgrade dovecot from version 2.3.15 to 2.3.16 for alpine >>>> linux. Build and 'make check' pass on all alpine architectures except on >>>> s390x which is big-endian while all other are little-endian. >>> >>> You're probably hitting the same issues like me on openSUSE. >>> >>> See Timo's response to that: >>> >>> https://dovecot.org/pipermail/dovecot/2021-August/122787.html >> >> The attached patch probably helps? > > Assuming I've applied the patch correctly it does not help on ppc64 and > S/390: > > https://build.opensuse.org/package/show/home:stroeder:network/dovecot23 <https://build.opensuse.org/package/show/home:stroeder:network/dovecot23>Well, that's annoying. I thought it would have worked. I wonder if it's still some issue with the #if not being right. Can you try once more with disabling the #if lines, i.e. just: +static void +copy_to_buf_last_used(struct mail_cache *cache, buffer_t *dest, bool add_new) +{ + size_t offset = offsetof(struct mail_cache_field, last_used); +//#if defined(WORDS_BIGENDIAN) && TIME_T_BITS > 32 + offset += sizeof(uint32_t); +//#endif + copy_to_buf(cache, dest, add_new, offset, sizeof(uint32_t)); +} If that doesn't work either, I'd somehow need to get access to a big-endian CPU to try to figure out where it's going wrong. -------------- next part -------------- An HTML attachment was scrubbed... URL: <https://dovecot.org/pipermail/dovecot/attachments/20210810/78d6ff2f/attachment.html>
Michael Ströder
2021-Aug-10 17:54 UTC
dovecot 2.3.16 fail one test on s390x, alpine linux
Timo, On 8/10/21 5:55 PM, Timo Sirainen wrote:> Well, that's annoying. I thought it would have worked. I wonder if it's > still some issue with the #if not being right. Can you try once more > with disabling the #if lines, i.e. just:It's currently still building on other big-endian platforms. But at least S/390 succeeded. (All low-endian platforms now fail but that's expected.)> If that doesn't work either, I'd somehow need to get access to a > big-endian CPU to try to figure out where it's going wrong.openSUSE build system is open for the public: Get yourself an account by clicking "Sign Up" here: https://build.opensuse.org You then have a home project home:username in which you can build any packages you want for a wide range of platforms. The build workers are often native hardware. You could branch the existing official devel dovecot23 package and do whatever tests you want to do: https://build.opensuse.org/package/show/server:mail/dovecot23 My own package builds are also just a branch of the above where I prepare the 2.3.16 update: https://build.opensuse.org/package/show/home:stroeder:network/dovecot23 Let me know if you need further information, though I'm not an OBS expert. Ciao, Michael.