reintroduce alloc_re() #159
Conversation
There was a problem hiding this comment.
I’m glad you found an example of a patch relying on this API (sooner rather than later) and are looking to keep such patches working while also communicating that they need updating (or merging).
In retrospect, I missed acting out my usual role here. Let me belatedly do the job: I’m worried that this C API transition is still too abrupt. Maybe, if we issue 1.09 with documented intent to deprecate alloc_re, users will tell us some other surprising reasons they’re depending on it before we decide to leave it out by default. And maybe not, of course. But I think it would be consistent with how we’ve been going about removing user-visible binaries. So I think I’m requesting that we leave alloc_re on-by-default for 1.09.
Tangent: assuming you mean this one, should we merge it? It’s been unconditionally applied in pkgsrc qmail for a long time. |
|
Yes, that one. Merging this should be considered. |
|
I reverted the #ifdef: it is now provided by default, but we can switch it off. I'm doing this in one CI build so we can detect if anything we add later would use it again. |
74dfd4a
to
8ede633
Compare
This is used by the large-dns patch, and may be used by others. Add it back in a way that it will not be compiled in our normal build, but a simple compiler define allows external patches to find the needed environment.
This is used by the large-dns patch, and may be used by others. Add it back in a way that it will not be compiled in our normal build, but a simple compiler define allows external patches to find the needed environment.