Learn more about these different git repos.
Other Git URLs
a bit late, I know, but going for it anyway...
F10 currently has e2fsprogs-1.41.0; I'd like to include 1.41.1
1.41.1 fixes a few things: yaboot FTBFS against 1.41.0, for one. Also e2image has a fairly severe bug:
{{{ E2fsprogs 1.41.0 intrduced a bug in libext2fs which casued e2image and debugfs programs to not be able to read e2image files; the signed vs. unsigned bug in the code which read bitmaps from the e2image has been fixed. }}}
there have also been several ext4-related fixes, such as resize support and more e2fsck fixes for ext4, and we'd like ext4 support to be as good as possible in this release.
there are 2 fairly major behavior changes, though:
I would revert these changes, or restrict them to ext4, so that we don't get these behavior changes on the default filesystem at the last minute.
so ... vote away ... or I'll happily answer any questions about this, first ;)
Thanks, -Eric
Do these changes effect people who are already installed, or is it just for new installs?
The journal placement & hash changes would affect fresh mkfs's only (if that's what you're asking?)
Ok, I'm reasonably comfortable with this then, plenty of snapshots before final.
+1
(Note: I don't think we need +1/+2 in this case, but thanks for bringing this up...)
Would the hash change make the filesystem not mountable on older releases/with older tools? If so, that's definitely a reason to restrict it to ext4.
Bill, thanks. I wasn't sure but wanted to err on the side of caution.
Actually no, the hash change should be backwards compatible.
I'm still inclined to not change it at this point for ext3 though. It's easy to restrict the default value change to ext4 in e2fsprogs, I have the patch ready to do so. It is also configurable via mke2fs.conf, tune2fs, and mkfs.
There have been some reports of more hash collisions w/ TEA and to add insult to injury, ext3 kernelspace does not cope with it well. But that bug has been there since forever and it's rarely hit.
Mulling this over to see whether I really want to push this change in at this point.
(Oh, and btw: e2fsprogs-1.41.2 just came out :) but it looks very safe, compared to the more functional changes that were in 1.41.1)
After talking a bit more with Jesse, I'm actually inclined to go ahead with the default hash change (which may actually require a vote, Jesse already did a +1) but patch out the journal move. The hash change arguably fixes a bug (since kernelspace isn't coping with collisions, and the old hash often yields collisions) but the journal move hasn't been shown to hae much benefit.
So if anyone wants to add another +1, I'll do the update with the hash change.
If this isn't desirable, I'll patch out any and all ext3 behavior changes, but still update for a few bugfixes.
+1 from me.
Eric, you're good to go, I'm closing this ticket.
Login to comment on this ticket.