Unknown branch format if line endings are modified

[Jaap Karssenberg]

Running Bazaar 0.91 on both Linux and Windows. Linux install came from distro, Windows install from installer on the website. Created and used branches on both hosts.

Gzipped a branch on Linux and copied it to the Windows host. Bzr is unable to use this branch:

  bzr: ERROR: Unknown branch format: 'Bazaar-NG meta directory, format 1\r\n'

Unix file endings are not the problem - converted line endings with gvim (on the Win32 host) - same error.

Copied to ".bzr\branch-format" file of a working Windows tree. Now the new error is:

  bzr: ERROR: Unknown branch format: 'Bazaar Branch Format 6 (bzr 0.15)\r\n'

Strangest thing is that I don't see a diff between the two "branch-format" files, but putting back the old one makes the first error return !?

Please suggest further tests - can submit gzipped branch that causes the error.

[Jaap Karssenberg (jaap.karssenberg)]

My bad. Line ending *is* the cause for the first error. Maybe line ending in another file is the cause for the second error?

Windows unzip utilities have a habit of converting to windows line endings :(

[John A Meinel (jameinel)]

Line endings are the problem.

And if your utility is changing all of the line endings, it is potentially corrupting other files. We store most of your data in compressed form. But that means that you might randomly have "\n" characters in those files. Which means they would be randomly expanded to "\r\n".

I would highly recommend either just using Bazaar itself to create the local branches, or just using a different decompression tool. (Maybe there is a flag to tell it not to break all of your files?)

We *could* be more lenient about the format strings. But I think in this case, it is better to detect that there has been modification, and just give a warning about possible corruption.

[Alexander Belchenko (bialix)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

John A Meinel пишет:
> Line endings are the problem.
>
> And if your utility is changing all of the line endings, it is
> potentially corrupting other files. We store most of your data in
> compressed form. But that means that you might randomly have "\n"
> characters in those files. Which means they would be randomly expanded
> to "\r\n".
>
> I would highly recommend either just using Bazaar itself to create the
> local branches, or just using a different decompression tool. (Maybe
> there is a flag to tell it not to break all of your files?)

I'm suggest to use 7-zip. It's open source, understand many formats,
and never touch line-endings as I can see.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHKLK2zYr338mxwCURAl1zAKCSeNk2NR66r6dJL5VddV1fHW4BAgCgikE6
OHWV3neLPFMB0NZrAflc/bc=
=ce03
-----END PGP SIGNATURE-----

[John A Meinel (jameinel)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Alexander Belchenko wrote:
> John A Meinel ?8H5B:
>> Line endings are the problem.
>
>> And if your utility is changing all of the line endings, it is
>> potentially corrupting other files. We store most of your data in
>> compressed form. But that means that you might randomly have "\n"
>> characters in those files. Which means they would be randomly expanded
>> to "\r\n".
>
>> I would highly recommend either just using Bazaar itself to create the
>> local branches, or just using a different decompression tool. (Maybe
>> there is a flag to tell it not to break all of your files?)
>
> I'm suggest to use 7-zip. It's open source, understand many formats,
> and never touch line-endings as I can see.

Agreed. The only thing I miss with 7-zip is simple drag-and-drop to Explorer.
But otherwise it has a huge supported set of formats and is free.

http://www.7zip.org

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHKLXaJdeBCYSNAAMRAl/1AKC4lvHnSr9COrw0yk/wE952amcaOQCg0+ml
rE6ythNKRAnD3Fe3ZtciLC8=
=Xga5
-----END PGP SIGNATURE-----

[Martin Pool (mbp)]

> We *could* be more lenient about the format strings. But I think in this
> case, it is better to detect that there has been modification, and just
> give a warning about possible corruption.

To be nice, we could give a specific message when the format string is
the same as a valid one, but with line-endings changed... (bug 159089)

--
Martin

[Jaap Karssenberg (jaap.karssenberg)]

Attached a small patch to throw a more descriptive error when line ending is wrong.

Disclaimer: I hardly know any python and patch is untested - consider it pseudo code

[Martin Pool (mbp)]

That looks reasonable. We should add a test that it works, by making
a repository and mangling its format.

Our first patch from cpan.org I think ;-)