xfsprogs process question
Hi Eric and others,
I've recently tried to port my attr cleanups to xfsprogs, but noticed that a lot of patches from others are missing before I could apply my ptches. Any chance we could try to have a real xfsprogs for-next branch that everyone can submit their ports to in a timely fashion? Without that I'm not sure how to handle the porting in a distributed way.
Quoted 8 lines · Christoph Hellwig
Hi Eric and others,
I've recently tried to port my attr cleanups to xfsprogs, but noticed that a lot of patches from others are missing before I could apply my ptches. Any chance we could try to have a real xfsprogs for-next branch that everyone can submit their ports to in a timely fashion? Without that I'm not sure how to handle the porting in a distributed way.
I guess the problem is that libxfs/ is behind, right. I have indeed always been late with that but it's mostly only affected me so far.
Would it help to open a libxfs-sync-$VERSION branch as soon as the kernel starts on a new version?
I've never quite understood what the common expectation for a "for-next" branch behavior is, though I recognize that my use of it right now is a bit unconventional.
-Eric
Quoted 13 lines · Eric Sandeen
Quoted 8 lines · Christoph Hellwig
Hi Eric and others,
I've recently tried to port my attr cleanups to xfsprogs, but noticed that a lot of patches from others are missing before I could apply my ptches. Any chance we could try to have a real xfsprogs for-next branch that everyone can submit their ports to in a timely fashion? Without that I'm not sure how to handle the porting in a distributed way.
I guess the problem is that libxfs/ is behind, right. I have indeed always been late with that but it's mostly only affected me so far.
Quoted 8 lines
Would it help to open a libxfs-sync-$VERSION branch as soon as the kernel starts on a new version?
I've never quite understood what the common expectation for a "for-next" branch behavior is, though I recognize that my use of it right now is a bit unconventional.
-Eric
I've run across this a few times working on the delayed attr series too. Sometimes I'll just have to go through the kernel side patches and find the missing pieces and port them myself just to get things to seat correctly. Eventually the xfsprogs side catches up, but it would seem to me that if we all did this it would help to catch up the user space side and keep it maintained. I've often wondered how one person manages to keep up with that! :-)
Allison
Quoted 1 line · Allison Collins
I've often wondered how one person manages to keep up with that! :-)
It seems that the answer is "badly" ;)
(TBH Darrick is a huge help here, in the long run.)
-Eric
Quoted 9 lines · Eric Sandeen
Quoted 1 line · Allison Collins
I've often wondered how one person manages to keep up with that! :-)
It seems that the answer is "badly" ;)
(TBH Darrick is a huge help here, in the long run.)
-Eric
:-) Wouldnt it make it easier though if people pushed a corresponding "xfsprogs: set name" that's already been applied/compiled and (hopefully) tested? That way when the user space side needs updating, patches can just be applied without one person having to work through all the ports? And then people could sort of cherry pick what they need off the list while the user space side is still between cycles. I realize this would generate a lot more traffic on the mailing list, but it seems like a reasonable way that people could help address this bottle neck.
I think I must have redone the userspace side of the delayed attr series over a dozen times, and there's a few patches that are not perfect mirrors of their kernel space counter parts. Because some changes affect code not present in the kernel space side, now all that needs updating too for the set as a whole to really work. It kinda seems like submitting just the kernel space side is still only half the set. Not sure how many other people run into that, but I suspect I'm likely not the first?
Allison
Quoted 2 lines · Eric Sandeen
Would it help to open a libxfs-sync-$VERSION branch as soon as the kernel starts on a new version?
Yes. I'd just call it for-next, but the name is the least of our problems :)
Quoted 3 lines
I've never quite understood what the common expectation for a "for-next" branch behavior is, though I recognize that my use of it right now is a bit unconventional.
Іt really is just a kernel convention due to linux-next. And I'd just name the xfsprogs branch to match. But as said above, the name isn't really what matters.