Attached, and the ebuild will require updating.
Created attachment 229673 [details, diff] new git freemint patch
alan, is this a patch suitable for git upstream? have you proposed it to them on the git mailing list?
No, it's not suitable for upstreaming yet and why it hasn't been done. There's a couple more tweaks to do, but this patch gets things compiling again.
Thing is without you taking this upstream such a patch may be needed on each version bump and we may need you again to keep the patch moving with releases. So it increases maintenance load downstream and delays bumps, something not in your interest. Please make it fit for upstream and have them apply it there. In case you care a single commit can get your name on <http://git-scm.com/about>.
(In reply to comment #3) > No, it's not suitable for upstreaming yet and why it hasn't been done. > > There's a couple more tweaks to do, but this patch gets things compiling again. > Yes, yes. Thanks. I'll add it in tomorrow if no one else does. To get to Sebastian's point, you know that we need patches submitted upstream as soon as possible to allow easy inclusion into gentoo-x86. With that being said, we have some leeway in our tree so it isn't soo strict. We have talked about this before, so it is nothing new. As more and more packages get moved to gentoo-x86, this requirement will get more strict because it is out of our control and moved to the individual Gentoo maintainers. Anyway, I hope that is clear to all as Sebastian has not been involved with your previous work and might not have known about your valued contributions and you might not have talked to him before now. Sebastian, thanks for the reinforcement and I agree with your statements on this bug, they are true.
Yep, completely understand. Feel free to leave this bug open until I get a new patch in that I'll submit upstream.
(gentoo git maintainer here) changing S_IFGITLINK is going to break stuff I'm pretty sure. Why was that being changed? Also, if you look at the existing Makefile, you'll see how to insert platform-specific stuff.
I'm 100% sure that the S_IFGITLINK will break, and that's why it's only applied when CHOST is mint for that reason. And I've mentioned it needs cleanup before submitting upstream and to feel free to leave this open until I've fixed things for inclusion upstream.
Just to mention on the S_IFGITLINK problem..... If you read the text before the define, the mechanism used is a bit of a kludge as they rely on the fact that 0160000 is S_IFDIR | S_IFLINK. But on MiNT that's not the case. S_IFDIR is 0040000 and S_IFLNK is 0160000, but S_IFMT is 0170000, so it's pretty broken here. I'm not sure how to fix that yet.
(In reply to comment #8) > I'm 100% sure that the S_IFGITLINK will break, and that's why it's only applied > when CHOST is mint for that reason. Actually I meant that I'm pretty sure any commits that you generate on MiNT (with that patch) containing a link will not be compatible with Git on other systems. And likewise repos from elsewhere containing a link probably won't work right on MiNT. I see you've taken it up with Junio upstream, so hopefully he has some success in his translation layer idea. Prefix: I suggest closing as RESO UPSTREAM for the moment and marking all of git with a -m68k-mint keyword, and then cherry-pick Junio's patchseries for the problem later.
I'm overwhelmed by the sudden attention to this bug. Alan, I'm confident you will act like you always do and end up with some patch which is acceptable for upstream. Maybe if there's interest from the git maintainers, we should put some efforts in merging git from the Prefix overlay back to gx86 now.
grobian: I spotted it while reviewing for Git bugs before doing the 1.7.1 version bump. Re other prefix changes, please open bugs for them, and I'll review to ensure that there isn't going to be breakage like this patch would have introduced.
It wouldn't have broken anything other than the FreeMiNT build as the ebuild only applies this patch in the CHOST == mint case. Anyway, I'm going to close this as UPSTREAM.
alanh: Err, specifically, I'm fairly certain that the S_IFGITLINK value is used in the index file and the tree objects, which do get propagated between machines. So commits generated on mint would be different in a subtle way than those on other platforms, and not compatible.
I know what you are saying......
