08:30 <haesbaert> apache2: no, long long >= long >= int

08:31 <haesbaert> they can all be 16bits for all it's worth (according to the spec)

08:31 <haesbaert> but the rule is, long is always 64 bits on 64bit unix, but still 32bit on windows

08:32 <haesbaert> my point is that long == int is perfectly valid, being the cause on all unix i386

12:14 <apache2> haesbaert: I'm with you on long == int being valid, that's defined by the platform spec, that's called IL32

12:15 <apache2> https://en.wikipedia.org/wiki/C_data_types#Basic_types

12:16 <apache2> this seems to suggest that the long long >= 64 bit is a C99 addition (potentially where long long was added?), but the long >= 32 bits and int >= 16 bits seems to be C89

12:18 <apache2> on linux and freebsd you have I32LP64, if that's the case for openbsd too then I don't see why this code needs that ifdef

12:19 <apache2> (I (and the code) assuming we're talking x86_64/amd64)

12:20 <apache2> so C spec-wise they can all be 64bits, but int is allowed to be smaller than long, which in turn is allowed to be smaller than long long -- whether that's the case is up to the compiler target

12:21 <apache2> in any case I think the way forward is for someone to install openbsd and take this for a spin and see where it breaks, if we are to continue to claim openbsd compatibility :)

12:22 <apache2> and my feeling is that we'll have to modify the templating for config.h to respect the stdint.h

12:26 <apache2> I looked up openbsd x86_64, they have long as 64-bit (as you said, and as reported by hystryfe), so that's not the problem here.

12:27 <apache2> (the conflicting type definition is, not sure why the special case is needed, but it looks like the CC doesn't equate the two types despite them being physically equivalent)

12:27 <apache2> so it's likely a partial work-around around that

16:17 <mort___> hannes: (or anyone really) can you recall which the cstruct release was that had the major api changes in? (around memory ownership etc) i have this vague recollection there was one, but can't immediately spot which in the changelog...

16:23 <hannes> mort___: cstruct 5.0.0 contained a new cstruct_cap module with ownership.

16:24 <hannes> afaict this is not (yet!?) used by anyone...

16:29 <mort___> i thought there was something prior to that (perhaps by several months) - a number of the function apis changed (extra parameters and so forht)

16:29 <mort___> perhaps i'm misremembering

16:30 <mort___> (we're trying to get the databox ocaml components updated to not be pinned to old releases (libraries and compilers) any more, one of which handles the networking)

22:42 <hystryfe> very interesting discussion :) I did get ocaml-freestanding to build and install by changing to solo5-bindings-hvt/crt/stdint.h: typedef long int64t; typdef unsigned long uint64_t; / and then commenting out the openbsd ifdef in config.in/m.x86_64.h.

22:43 <hystryfe> However... mirage-solo5 looks like it has the same problem. while probably not the correct solution, I just want it to work to see what I can do with it.

22:44 <hystryfe> (same problem -> i'll attempt the same solution)

23:42 <hystryfe> or... not same error, and I will have to do it the correct way. I'm just fiddling around because another issue I'm having is setting up a share to get logs to my host. Don't even think of samba, and nfs has got a case of the RPCs. Will continue in a bit. Work called. Keep pinging out. Thanks all.