[PD] Miller's pd 0.40 OSX loader

IOhannes m zmoelnig zmoelnig at iem.at
Thu Sep 21 22:56:16 CEST 2006


james tittle wrote:
> On Sep 21, 2006, at 3:17 PM, IOhannes m zmoelnig wrote:
>>
>> wow! do the change the preferred way of linking with every minor release?
>>
>> i can understand that there might be some changes in the preferred
>> loading mechanism after the macintel change (with the goal of supporting
>> both architectures).
>> but afair, they only changed the fundamental architecture once in the
>> last 10 years...
> 
> ...well, to be fair, the switch was done with 10.3.x, which is about 4
> years ago at this point...and look at this as a good thing, because I
> remember everyone complaining in the pre-10.3 days about the lack of
> dlopen() standardizing...

so it's just the pd-community that is lagging somewhat behind?

> 
>> so should we change the configure-checks for Gem?
>> to either use "-dynamiclib -mmacosx-version-min=10.3 -undefined
>> dynamic_lookup" or - if this is not supported - "-bundle -undefined
>> suppress -flat_namespace"
> 
> ...it depends:  I wouldn't do the check fully based on osx version, but
> perhaps also include pd version, since I don't recall if the older
> NSBundle stuff works with "dynamiclibs" or not...

but that pretty soon becomes ugly.
afaik, dlopen() is the standard way of loading externals on osx at least
since pd-0.39.
since we are rapidely approaching pd-0.40 i guess we could just go and
ignore previous versions.
as for the os-x version: isn't 10.3 the minimum requirement for pd anyhow?

so the check as i currently do it is:
"does the compiler support "-dynamiclib -mmacosx-version-min=10.3
-undefined dynamic_lookup"? if not, does it support "-bundle -undefined
suppress -flat_namespace"?
i hope this will just work in most cases.

> 
> ...always keeping us on our toes...

good to see you back on yours: getting more lively, are ye ;-)



,fgasdr
IOhannes





More information about the Pd-list mailing list