Re: Link failures of ArcticFox on PPC

Riccardo Mottola <[email protected]> Mon, 22 Apr 2024 12:56:28 +0200
Newsgroups gmane.os.netbsd.ports.macppc
Message-ID <[email protected]>
Hi Martin

Martin Husemann wrote:
> Are you building from pkgsrc or the upstream code? 

I am building upstream code directly. "I am upstream" :)

> Check wich of the  xpcom/reflect/xptcall/md/unix/xptcstubs_ppc_* files
> gets used in your build, it could be xptcstubs_ppc_netbsd.cpp instead
> of xptcstubs_ppc_linux.cpp (for probably no good reason but history).
>
> And if it is the netbsd variant, it probably may not have been updated
> when some other sweep was done.

Thank you for the hint. I have checked and there aren't many updates on 
that from Firefox up to 52, which probably doesn't care much about PPC 
and even less NetBSD.

You brought up a PR though which I never admitted because I didn't fully 
understand and I want more on these delicate things:

https://github.com/rmottola/Arctic-Fox/pull/101/commits

I did try to get more information, but the requester was very... short.

I dug it up and tried to apply two of his commits.

As is:
https://github.com/rmottola/Arctic-Fox/pull/101/commits/6a3f3104b677779c63f37f228b3cc9833af6bc33

For
https://github.com/rmottola/Arctic-Fox/pull/101/commits/19a4aac9b7dc082c5c0c6973c374cc5fc9acf662
I took a different approach, I copied the OpenBSD files over the NetBSD 
files, should be the same, but liked that more

I ignored this part of the patch:
https://github.com/rmottola/Arctic-Fox/pull/101/commits/949663ff28387756e89bbc1196cd47e2697afda3#diff-8d368f59e632f6fcd46a6e3ef233c2ec71db8ccf4b45b73f70c0f0682c8f6a75

couldn't find a trace for it in Gecko and it shouldn't directly affect 
linking, from what I understand.

Riccardo

PS: I have then an additional problem. Sources should rebuild 
incrementally - like expected from a make system. I know it is a hybrid 
python/make thing made custom. It happens for me that "./mach build" 
every time appears to do a full build - about 10h of compilation. It 
happens (I don't remember if always or sometimes( to me also on 
NetBSD/sparc64 (20hours there...). It doesn't happen on NetBSD intel, be 
it 32bit or 64bit. Nor does it happen on Linux/PPC.... it makes things 
really hard to test, as you imagine. Poor iBook used as a compilation 
farm :)