Skip to contentSkip to conversation

Quilt 0.43 has been released! [SERIOUS BUG]

Quilt-dev Peter Williams 4 participants 7 messages Feb 14, 2006 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw
Quoted 7 lines · Jean Delvare

Hi all,

Quilt 0.43 is now available for download from Savannah:
http://download.savannah.gnu.org/releases/quilt/

Enjoy!

I've experienced a serious problem with this version of quilt that is rather hard to explain but definitely arrived with this version as reverting to 0.42 makes the problem go away.

Now to try to explain the problem :-(

The problem arises when pushing a patch that has errors in it (due to changes in the previous patches in the series) and needs the -f flag to force the push. What's happening is that the reverse of the errors is being applied to the "pre patch" file in the .pc directory. Then when you pop this patch it returns the file to a state with the reverse of the errors applied to it.

I'm having trouble understanding how quilt could be dumb enough to do this as surely the "pre patch" file in the .pc directory should be just a copy of the file before the patch is applied.

This bug can completely hose a set of patches if the user doesn't notice it very early and do something about it. The work around is to revert to version 0.42 of quilt.

Peter

Signature · 4 lines
Peter Williams                                   [email protected]

"Learning, n. The kind of ignorance distinguishing the studious."
 -- Ambrose Bierce
#2 lore raw
Quoted 6 lines · Peter Williams

The problem arises when pushing a patch that has errors in it (due to changes in the previous patches in the series) and needs the -f flag to force the push. What's happening is that the reverse of the errors is being applied to the "pre patch" file in the .pc directory. Then when you pop this patch it returns the file to a state with the reverse of the errors applied to it.

Found and fixed. It's a missed rollback_patch on one of the two branches of the code that checks if a patch can be reverse applied. This case apparently doesn't trigger as easily as it seems, or else we would have found it sooner. Still quite bad.

Shall we wait until the translations are up-to-date again, or release 0.44 immediately?

Quoted 3 lines

I'm having trouble understanding how quilt could be dumb enough to do this as surely the "pre patch" file in the .pc directory should be just a copy of the file before the patch is applied.

Hey, it's just a bug.

Quoted 3 lines

This bug can completely hose a set of patches if the user doesn't notice it very early and do something about it. The work around is to revert to version 0.42 of quilt.

You should have sent your gquilt announcement to the quilt-dev list as well. Thank you for summarizing the changes.

Andreas

#3 lore raw
Quoted 11 lines · Andreas Gruenbacher
Quoted 6 lines · Peter Williams

The problem arises when pushing a patch that has errors in it (due to changes in the previous patches in the series) and needs the -f flag to force the push. What's happening is that the reverse of the errors is being applied to the "pre patch" file in the .pc directory. Then when you pop this patch it returns the file to a state with the reverse of the errors applied to it.

Found and fixed. It's a missed rollback_patch on one of the two branches of the code that checks if a patch can be reverse applied.

That was quick.

Quoted 11 lines

This case apparently doesn't trigger as easily as it seems, or else we would have found it sooner. Still quite bad.

Shall we wait until the translations are up-to-date again, or release 0.44 immediately?

Quoted 3 lines

I'm having trouble understanding how quilt could be dumb enough to do this as surely the "pre patch" file in the .pc directory should be just a copy of the file before the patch is applied.

Hey, it's just a bug.

:-)

Quoted 8 lines
Quoted 5 lines

This bug can completely hose a set of patches if the user doesn't notice it very early and do something about it. The work around is to revert to version 0.42 of quilt.

You should have sent your gquilt announcement to the quilt-dev list as well. Thank you for summarizing the changes.

Sorry. They were mainly due to your change from "-p <patch>" to "-P <patch>" for some commands. The other issue was a change to the return value and error message when "quilt top" was used in a directory without any quilt data. So no real changes to gquilt as seen by the user just implementation changes.

BTW the --version function made it possible to make gquilt still work with versions earlier than 0.43.

Thanks
Peter

Signature · 4 lines
Peter Williams                                   [email protected]

"Learning, n. The kind of ignorance distinguishing the studious."
 -- Ambrose Bierce
#4 lore raw
In reply to Andreas Gruenbacher

Hi all,

Quoted 8 lines · Andreas Gruenbacher
Quoted 6 lines · Peter Williams

The problem arises when pushing a patch that has errors in it (due to changes in the previous patches in the series) and needs the -f flag to force the push. What's happening is that the reverse of the errors is being applied to the "pre patch" file in the .pc directory. Then when you pop this patch it returns the file to a state with the reverse of the errors applied to it.

Quoted 4 lines

Found and fixed. It's a missed rollback_patch on one of the two branches of the code that checks if a patch can be reverse applied. This case apparently doesn't trigger as easily as it seems, or else we would have found it sooner. Still quite bad.

I probably encountered it the other day, but as I couldn't explain what was happening, I mistakenly concluded to a user error and started again from a fresh tree. Or maybe it was really a user error after all.

I was about to suggest that we add a test in the quilt test suite, but I see you did already - good!

Quoted 2 lines

Shall we wait until the translations are up-to-date again, or release 0.44 immediately?

I'd say:
* Fix the temporary file leak in the mail command I have been reporting a few days ago - unless it's there on purpose.
* Update the translations. I'll take care of French this evening (GMT+01).
* Let people (including me) do a little testing. If nothing else, running the test suite on a few different systems can't hurt.
* Release.

We can be done by tomorrow if Yashi can handle the Japanese translation fast. If Yashi is too busy I guess we'll have to release anyway...

Thanks,
--
Jean Delvare

#5 lore raw

Hi Jean,

At Tue, 14 Feb 2006 09:18:39 +0100 (CET),

Quoted 13 lines · Jean Delvare
Quoted 3 lines

Shall we wait until the translations are up-to-date again, or release 0.44 immediately?

I'd say:
* Fix the temporary file leak in the mail command I have been reporting a few days ago - unless it's there on purpose.
* Update the translations. I'll take care of French this evening (GMT+01).
* Let people (including me) do a little testing. If nothing else, running the test suite on a few different systems can't hurt.
* Release.

Quoted 2 lines

We can be done by tomorrow if Yashi can handle the Japanese translation fast. If Yashi is too busy I guess we'll have to release anyway...

:-) here is my part.
--
yashi

#7 lore raw
In reply to Jean Delvare
Quoted 19 lines · Jean Delvare

Hi all,

Quoted 9 lines · Andreas Gruenbacher
Quoted 6 lines · Peter Williams

The problem arises when pushing a patch that has errors in it (due to changes in the previous patches in the series) and needs the -f flag to force the push. What's happening is that the reverse of the errors is being applied to the "pre patch" file in the .pc directory. Then when you pop this patch it returns the file to a state with the reverse of the errors applied to it.

Quoted 4 lines

Found and fixed. It's a missed rollback_patch on one of the two branches of the code that checks if a patch can be reverse applied. This case apparently doesn't trigger as easily as it seems, or else we would have found it sooner. Still quite bad.

Quoted 3 lines

I probably encountered it the other day, but as I couldn't explain what was happening, I mistakenly concluded to a user error and started again from a fresh tree. Or maybe it was really a user error after all.

I went through much the same process (several times) before it finally dawned on me that it might be a problem with the newly updated quilt and replaced it with 0.42 :-(. I was worried that the symptoms were so bizarre and hard to describe properly that the problem would be hard to fix so I was impressed by the speed with which the problem was solved.

Quoted 18 lines

I was about to suggest that we add a test in the quilt test suite, but I see you did already - good!

Quoted 2 lines

Shall we wait until the translations are up-to-date again, or release 0.44 immediately?

I'd say:
* Fix the temporary file leak in the mail command I have been reporting a few days ago - unless it's there on purpose.
* Update the translations. I'll take care of French this evening (GMT+01).
* Let people (including me) do a little testing. If nothing else, running the test suite on a few different systems can't hurt.
* Release.

Quoted 2 lines

We can be done by tomorrow if Yashi can handle the Japanese translation fast. If Yashi is too busy I guess we'll have to release anyway...

Peter

Signature · 4 lines
Peter Williams                                   [email protected]

"Learning, n. The kind of ignorance distinguishing the studious."
 -- Ambrose Bierce