| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Sometimes the tar files will contain PAX extended headers to deal with
cases where the information needed doesn't fit initial header.
One of these cases is when the path is larger than 100 characters. An
extended header will appear before the block including the contents of
this file, including information about the entire path.
The PAX extended headers contain one or multiple records constructed as
follows:
"%d %s=%s\n", <length>, <keyword>, <value>
This commit makes sure that we always read the extended header blocks,
and in the case of finding one, we parse its records looking for 'path'
information. If this information is found, is stored for the next
iteration.
As a side effect, this commit also fixes a bug where we weren't ignoring
the extended header contents because the truncated path was ending on '/'
and the script was skipping any actions given that it thought that it
was a folder (and it wasn't).
|
| |
|
|
|
|
|
|
|
| |
If parsing the lorry as YAML fails, we attempt to parse it as JSON
before giving up completely.
Change-Id: I7e3502f5ae9c32e10d63047e5f280bfe1d7e52dc
|
|
|
|
|
|
|
|
| |
This prevent git from running processes in the background. We were
having issues when `git gc` was running automatically, making `git gc`
call fail because it was already running.
Change-Id: Iee3e8e29eb0697f5241c7de4744c8e080bfd8e28
|
|
|
|
|
|
|
|
| |
We can do everything we need to on the command-line,
which makes more sense than syncing lorry file config every time,
and it makes it easier to add extra options later.
Change-Id: Ieaabc9fe7a83b5c29371ddb92b991d128c90e6c6
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
HG lorries are currently broken with current versions of hg-fast-export
Filed upstream bug here because it technically should not be broken:
https://github.com/frej/fast-export/issues/57
Note the issue in the upstream git repository here:
http://repo.or.cz/fast-export.git/blob/refs/heads/master:/hg-fast-export.sh#l54
This patch works around the issue by simply reordering the arguments
of hg-fast-export in the lorry program.
Change-Id: I2c8b303e2d90edb301186c0b3a6c57ba2e0b8179
|
|
|
|
| |
Change-Id: Ifd8fd2c59f218a478ddfce67c60c623f77a69de8
|
|
|
|
|
|
|
|
| |
The new 'gzip' lorry file is practical for any gzipped single file
repositories, this can be useful for some static blob resources,
such as fonts.
Change-Id: I931e15e4c1291d820db82d0fc764aa767d3df9ab
|
|
|
|
| |
Change-Id: I55e6fd73d39b0cf0e20da27a99ddcced5b5b4063
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Some svn repos don't have a branches dir or a tags dir in their layout.
This will for example allow us to lorry xmlto with the following lorry,
{
"xmlto": {
"type": "svn",
"url": "http://svn.fedorahosted.org/svn/xmlto",
"layout": {
"trunk": "."
}
}
}
Change-Id: I74c032ab7642f4b22ed0569b2a163c0b78cba658
|
|
|
|
|
|
|
| |
Any extended headers present (such as those created by bsd tar)
should be ignored.
Change-Id: Ic1c0831c349fc7192892fbf487427dac3b320366
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
bsd tar stores file attributes as extended headers,
which are ignored by gnu tar. The tar importer currently
treats these headers as regular files, so regular files
containing attribute data are created on extraction even
though they never existed at archive creation.
This can be a problem for CPAN imports, some CPAN distributions
are created using bsd tar, the extended headers in these archives
result in a modified directory structure, e.g. where the structure
on the developer's machine would be
Foo:
META.json
META.yml
Makefile.PL
...
in trove we could have,
Foo:
Foo/
Foo/META.json
Foo/META.yml
Foo/Makefile.PL
PAXHeader/
PAXHeader/Foo
...
this structure causes the default build-commands to fail because they
expect to find Makefile.PL in the repo's root dir.
Change-Id: Ia8f90c3be7b31f7a4ac774022cb606fc1c57c002
|
|
|
|
| |
Change-Id: I8158e53ff64e7ed68e469971b15ba78b0d5fef1c
|
|
|
|
|
|
| |
For example, we'd rather the tag be docbook-xml-4.5 than docbook-xml-4.5.zip
Change-Id: I617b1c577ff95e4231681c75d3a91f4826cafd2b
|
|
|
|
|
|
|
|
|
| |
Include only the basename of the file, to make the commit message
(and thus commit SHA1s) deterministic.
Mention Lorry in the committer name and email address.
Change-Id: Id9d29bbe9f12514213b4b4eafd35b08b017e4246
|
|
|
|
| |
Change-Id: I558986188396981c1caf2adf19860c9fbed2f46f
|
|
|
|
|
|
|
| |
We don't want the sha to depend on where the lorry was imported from
or the time at which the lorry occurred, or the user's environment.
Change-Id: Iccc54f358a1a6133438984e769485b702ebb1a2e
|
|
|
|
|
|
| |
Also update copyright
Change-Id: Iff42a32048482520cee695bc1262990f5989ce63
|
|
|
|
| |
Change-Id: I29368ddcc64c82bafcbb834c81db30018a144f64
|
|
|
|
|
|
|
| |
Add zip archive frontend for git-fast-import from
https://raw.githubusercontent.com/git/git/master/contrib/fast-import/import-zips.py
Change-Id: I7cbd63bda5a1c36df68bef6448acbc55d95672b4
|
|
|
|
| |
Change-Id: I78eceb3100c8d1c84cf8f2aa929749cd4e33bd68
|
|
|
|
|
|
|
|
| |
Add a fake mirror-base-url-fetch for running make-tarball test.
This was broken because if mirror-base-url-fetch is empty
lorry avoids to create the bundle and the tar file.
Change-Id: I606226fb26739a366641b8f3ac9c9969860abcb0
|
|\
| |
| |
| | |
Reviewed-By: Sam Thursfield <sam.thursfield@codethink.co.uk>
|
|/ |
|
|\
| |
| |
| |
| | |
Reviewed-By: Pedro Alvarez <pedro.alvarez@codethink.co.uk>
Reviewed-By: Javier Jardón <javier.jardon@codethink.co.uk>
|
| |
| |
| |
| |
| | |
The 'compression' and 'strip' fields seem to have been obsolete since
commit a7f8b1b63a97c66aea6f39342313d7a5b919cfa7 in 2012.
|
|\ \
| |/
|/|
| |
| | |
Reviewed-By: Pedro Alvarez <pedro.alvarez@codethink.co.uk>
Reviewed-By: Adam Coldrick <adam.coldrick@codethink.co.uk>
|
| |
| |
| |
| |
| |
| |
| |
| | |
It doesn't make sense to have Lorry push to a random default mirror
server: users running Lorry without setting up any configuration are
likely to be testers/developers that don't want it to randomly push
stuff to places. It especially doesn't make sense to refer to the
Codethink 'roadtrain' server which was decommissioned several years ago.
|
|\ \
| |/
|/|
| |
| | |
Reviewed-By: Adam Coldrick <adam.coldrick@codethink.co.uk>
Reviewed-By: Francisco Redondo Marchena <francisco.marchena@codethink.co.uk>
|
| |
| |
| |
| |
| |
| | |
This is a safety measure. Commit
8bddb0996c44ce8ce989359a7b4c9ff102466ef2 should mean that this can't
happen, but 'lorry' runs unattended so we really don't want it to break.
|
|/
|
|
|
|
| |
Previously if a tarball lorry failed, a 0-byte tar file would be kept
around and subsequent calls to lorry would return success without
actually importing anything.
|
|\
| |
| |
| |
| | |
Reviewed-by: Richard Maw
Reviewed-by: Daniel Silverstone
|
| |
| |
| |
| |
| |
| | |
Similar to the url config change, if we need to change the refspecs
for which branches to pull, then we need to apply them at every run,
rather than just the first run.
|
| | |
|
|/ |
|
|\
| |
| |
| |
| |
| |
| |
| | |
Daniel and Richard reviewed the original set of changes, and Richard my
further changes based on review feedback.
Reviewed-by: Daniel Silverstone
Reviewed-by: Richard Maw
|
| | |
|
| | |
|
| | |
|
|/ |
|
| |
|
|\
| |
| |
| |
| | |
Reviewed by:
Daniel Silverstone
|
| | |
|
| | |
|
| | |
|
| | |
|
|/
|
|
|
|
|
| |
This test didn't have the correct expected output.
This commit corrects the expected output and amends
the test so that it fails if there are no pack files
in the test backup repo.
|
| |
|
| |
|
| |
|