[FEAT] Merge of pull request "Split pybind11 into headers and cpp files to speedup compilation #2445"

[gabemblack]

Hi, this may not be exactly the right place to ask this question, but could someone please tell me what the timeline is for merging the pull request mentioned in the title of this issue, "Split pybind11 into headers and cpp files to speedup compilation #2445"?

I commented on the pull request about a week ago and have not seen a response. It looks like, and from talking to @cirosantilli2 (the author) sounds like, there has been good support for this change which has gone through revision and was in line to merge, waiting for after an upcoming release window. This change would be very valuable to the gem5 simulator project and others I'm sure, by saving human hours and machine hours in the develop cycle, continuous integration services, periodic sweeps of build targets, etc.

I know that Ciro would like to get that code merged (as would I!), and it sounds like the pybind11 folks would also. What do we need to do to move that forward?

[EricCousineau-TRI]

Howdy @gabemblack! I'm checking w/ others to see if we should queue this up for v2.7 or another release!

[gabemblack]

Great, glad to hear it, thank you!

[gabemblack]

Hi Eric, is there any news as far as this patch merged? Thanks!

[gabemblack]

Bump...

[gabemblack]

Hi Eric, it's been a month since I last heard from you. Could you please let me know if anything is happening?

[EricCousineau-TRI]

Howdy Gabe, my deepest apologies. I don't think this would make it to the 2.7 release; however, it could still be considered for the future!

(FWIW I work on a project that could benefit from this; however, it's not yet in our critical path to motivate it just yet.)

[gabemblack]

Hi Eric, no worries, I'm not concerned about it making a particular release. I do really want to see it make it into a release however, hopefully in the not too distant future?

Again, if there's anything we can do to help this along, please let me know. This would be a big quality of life improvement for our project.

[gabemblack]

Hello again Eric, has there been any update on this issue? I heard through the grape vine that this would be discussed at a pybind11 meeting a while ago?

[gabemblack]

Ping!

[EricCousineau-TRI]

It was briefly discussed, with some notes captured here: https://github.com/pybind/pybind11/wiki/Roadmap#targets-for-28

I need to stay "shrouded" until mid-October; after which, I'll try to re-engage.

[gabemblack]

Hi Eric, I was happy to see that version 2.8 had been released, and was very much looking forward to using the "precompilation" feature that was on the roadmap and that we'd been discussing here. To my dismay, I see no mention of that feature in the changelog, nor do I see it in the copy of the code I just checked out.

Can you please tell me what happened, and when we might actually see this feature? This is such a pain point for us, we may have to consider forking pybind11 to bring our builds back under control.

[gabemblack]

Hi Eric, has there been any change on this front? Is there any chance of this feature getting into a future version? This would still be very valuable for us. It's very frustrating to have waited so long for this feature, and to see it hang on the edge of being implemented only to not happen. Believe me, I understand that there are always ten times more to do than people to do them, but I can say for us in particular, and probably for a lot of other projects in general, this would be a very welcome quality of life improvement when using pybind.