Skip to contentSkip to conversation

Unintended commit?

DISCUSSION Claudio Martins 5 participants 6 messages Oct 23, 2006 – Oct 24, 2006

lore
Share & tools
lore
#0 lore raw

Hi,

I noticed that commit 4e8a5201506423e0241202de1349422af4260296 on Linus' tree
titled "[PKT_SCHED] netem: Orphan SKB when adding to queue." also touches the file "drivers/pci/quirks.c", which seems unrelated.

Was this intentional?
If so, then I'm sorry for the noise...

Best regards

Claudio Martins

#1 lore raw
Quoted 4 lines · Claudio Martins

I noticed that commit 4e8a5201506423e0241202de1349422af4260296 on Linus' tree
titled "[PKT_SCHED] netem: Orphan SKB when adding to queue." also touches the file "drivers/pci/quirks.c", which seems unrelated.

Indeed. Thanks for noticing.

David, can you pls explain?

Linus

#2 lore raw

From: Linus Torvalds <[email protected]>
Date: Mon, 23 Oct 2006 10:13:52 -0700 (PDT)

Quoted 3 lines

Indeed. Thanks for noticing.

David, can you pls explain?

Totally my mistake.

A PCI change a few days ago blows up Sparc64 so I kept putting in that drivers/pci/quirks.c workaround in order for me to be able to test the current tree.

I accidently commited it, again sorry. Greg KH has a patch coming to you soon which will move that VGA code back into x86/x86_64/IA64 specific areas and will fix the sparc64 problem properly.

I should make some modifications to my workflow so that I don't do this again when I need to make local unrelated changes in order to test the current tree.

When I'm grinding out a patch actively in my tree I go "git diff >diff" and then I use a script called "mkcf" which runs diffstat on the diff and gives me a file list, it's ugly, but here it is:

#!/bin/sh
diffstat -p 1 $* | grep -v changed | awk ' { print $1 } '

So I end up going:

1) edit files
2) git diff >diff
3) read diff
4) git commit $(mkcf diff)

Usually step #3 catches local changes I'm not intending to commit and I just edit those out, and therefore they never end up being committed.

Perhaps it would be cool if you could tell GIT "Look, I know I have a change to foo.c, but it's a local hack and please act like it's not there unless I try to do an operation where ignoring that change is impossible, such as a merge."

The idea is that things like "git diff", "git update-index" etc. would not show the change, but things like a "git pull" that would reference the file with such changes would conflict.

#3 lore raw
Quoted 4 lines · David Miller

I accidently commited it, again sorry. Greg KH has a patch coming to you soon which will move that VGA code back into x86/x86_64/IA64 specific areas and will fix the sparc64 problem properly.

Ok. I reverted that part, we'll get it sorted out properly with Greg's patches.

Claudio, thanks for noticing!

Linus

#4 lore raw
In reply to David Miller
Quoted 19 lines

I should make some modifications to my workflow so that I don't do this again when I need to make local unrelated changes in order to test the current tree.

When I'm grinding out a patch actively in my tree I go "git diff >diff" and then I use a script called "mkcf" which runs diffstat on the diff and gives me a file list, it's ugly, but here it is:

#!/bin/sh
diffstat -p 1 $* | grep -v changed | awk ' { print $1 } '

So I end up going:

1) edit files
2) git diff >diff
3) read diff
4) git commit $(mkcf diff)

Quoted 8 lines

Usually step #3 catches local changes I'm not intending to commit and I just edit those out, and therefore they never end up being committed.

Perhaps it would be cool if you could tell GIT "Look, I know I have a change to foo.c, but it's a local hack and please act like it's not there unless I try to do an operation where ignoring that change is impossible, such as a merge."

I prefer to explicitly commit all changes, i.e.:

edit files
svn diff or svn diff | lsdiff or both
svn ci -m foo a/b.c c/d.c etc.

might be a little lot to type, but it usually helps committing the right stuff.

(Yes, it is intentional that I wrote svn instead of git.)

	-`J'
#5 lore raw
In reply to David Miller

Hi!

Quoted 4 lines

Perhaps it would be cool if you could tell GIT "Look, I know I have a change to foo.c, but it's a local hack and please act like it's not there unless I try to do an operation where ignoring that change is impossible, such as a merge."

When I was solving that problem, I ended up with two repositories:

clean-linus <-> linux-pavels-hacks <-> linux.

That behaves mostly as you want it...

Pavel

Signature · 1 line
Thanks for all the (sleeping) penguins.