| 1 |
I figure this is half-on-topic for this list since I'm trying to |
| 2 |
prepare patch sets for a package. I'm getting fairly bizarre behavior |
| 3 |
from git format-patch - patches that don't apply, and patches numbered |
| 4 |
early in sequence that didn't show up previously in this branch. I |
| 5 |
suspect rebasing might be the cause of that change, but I don't think |
| 6 |
it fully explains this behavior. |
| 7 |
|
| 8 |
To demonstrate, run: |
| 9 |
git clone https://github.com/MythTV/mythtv.git -b fixes/0.26 |
| 10 |
cd mythtv/ |
| 11 |
git format-patch v0.26.0 |
| 12 |
mv *.patch .. |
| 13 |
git checkout v0.26.0 |
| 14 |
patch -p0 < ../0001-* |
| 15 |
|
| 16 |
Final output is: |
| 17 |
can't find file to patch at input line 17 |
| 18 |
(messing with -p doesn't help, which will be obvious from a quick |
| 19 |
inspection of the file vs the tree) |
| 20 |
|
| 21 |
How can git format-patch against a tag generate a patch that won't |
| 22 |
apply against that tag? |
| 23 |
|
| 24 |
If you look at the git log of that branch you'll find that the first |
| 25 |
patch is from a commit that is more distant in the past than the tag. |
| 26 |
|
| 27 |
Am I relying on undefined behavior? This has generated useful patches |
| 28 |
for me in the past... |
| 29 |
|
| 30 |
Rich |
Archives