When using BB_GENERATE_MIRROR_TARBALLS="1" to generate mirror tarballs
of git repositories, they leaked local information: username, group and
time of the last fetch. Remove all these by setting fixed information:
* uname = pokybuild
* gname = users
* mtime = committer time of newest commit in repo
The username and group value were taken from the archives available on
the downloads.yoctoproject.org mirror. The modification time is chosen
so it still retains some relationship to the contents of the archive.
(Bitbake rev: 0178ab83e6312e97e528aa8c5e12105f5165d896)
Signed-off-by: Olaf Mandel <o.mandel@menlosystems.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Where a git url uses a tag instead of a full source revision, breakage
can currently occur in builds. Issues include:
* the revision being looked up in multiple tasks (fetch and unpack)
* the risk a different revision may be obtained in those tasks
* that some tasks may not be allowed to access the network
* that a revision may not be consistent throughout a given build
* rerunning a specific task may given inconsistent results
To fix this, stop the workers from cleaning out the source revision store. This
should only be done in the cooker itself (based on current policy).
Also, where the code "sees" an upstream access, mark the recipe as not to be
cached. The reparse re-triggers the upstream lookup by the server.
Add a test to ensure that if get_srcrev isn't called, the user is told they're
using a configuration that is known to break.
(Bitbake rev: 4b5eed1626709ef3dc06b32fd55d40a2a6edd179)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
On master, tell the users they need to update their urls for github.
(Bitbake rev: 42526a402357e04794f4cb6f21ac18f562220a9b)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
github is dropping support for git protocol in Git urls. Add code to remap
this to https in a way that could be used in older bitbake versions.
(Bitbake rev: f05e19135b3ddee509c0cb427b3b9376bb4738d1)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
There is much uncertainty around what tools and hosting providers will
do about default git branch naming in the future. To help ensure we
can handle the various scenarios, we will make branch names required in
SRC_URI. To start that process, show users a warning if it isn't set.
This may also allow us to change the default at some point in the future.
(Bitbake rev: 86a9c26828479be55865bcce72bcc7e12b93caa7)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The git fetcher is odd in that it supports destsuffix as a parameter but not
the default documented subdir parameter. destsuffix is more limited as it can't
take absolute paths. Rework the code to correctly support subdir.
Also cleanup to use the None default .get() values and be a bit more pythonic
and use subpath as the variable name for subpath for code clarity.
We could consider dropping destsuffix as a parameter as some future point.
Also fix the tests not to pass in a subdir parameter which was never used
but now causes errors.
(Bitbake rev: 66953f06fe822e4001efabd9fc1c985ea2b03f96)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
os.rename will overwrite the destination file if present so we can use this
instead of the process call overhead.
(Bitbake rev: b3cccaa6a896c41d8c9be5eebc327f726542d16b)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
There is a potential race over the mirror tarballs where a partial git repo
could be extracted causing fetcher failures if the tarball is being rewritten
whilst another build accesses it.
Create the mirror tarball atomically to avoid this.
[YOCTO #14441]
(Bitbake rev: 3250bc950c56bd7dd2114df26e5a8e13b04ceac8)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It looks like git gc can interrupts the package creation when
BB_GENERATE_MIRRORS_TARBALL is in use.
Log excerpts:
tar -czf TOPDIR/../downloads/git2_bitbucket.name-hidden.git.tar.gz . failed with exit code 1, output:
tar: ./objects/pack/pack-89a1d76f6c08f53172ef1d02ff851d90564362c4.pack: file changed as we read it
tar: ./objects/pack/pack-b4a48ada355d333630fdf6b4f67205b7c264dc2c.idx: File removed before we read it
Auto packing the repository in background for optimum performance.
See "git help gc" for manual housekeeping.
(Bitbake rev: a8d8cb847063862d1a7998963dd8b767ff73d877)
Signed-off-by: Adam Romanek <romanek.adam@gmail.com>
Signed-off-by: Damian Wrobel <dwrobel@ertelnet.rybnik.pl>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Gitsm fetcher crashes when cloning a repository that contains LFS files.
This happens because the unpack method is called during download, but the
submodules have not been downloaded yet at this point.
This issue was introduced in this
commit: 977b7268bf
[YOCTO #14283]
(Bitbake rev: 26caedc4d2e9b5a0f1d57f9291754a7f6c5e437e)
Signed-off-by: Niels Avonds <niels@codebits.be>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
By default the unpacker will create a "shared" clone when cloning from
the DL_DIR to the WORKDIR. This patch introduces an option to control
that behaviour.
Imagine some recipe steps are executed in a namespace that is different
from the one your downloader and unpacker ran in. (chroot) Because a
"shared" clone has an absolute reference to its "alternate" you now
have to make that "alternate" visible in that new namespace (chroot) at
the exact place.
With this patch you can unpack "noshared" and get a stand-alone copy.
This copy will also work if the "alternate" is not visible or existant.
The switch is a global bitbake switch and will affect all git urls.
Build systems that need "noshared" most likely need it for everything
they do with git.
(Bitbake rev: 6ae6f1865d5e666ebc670f70b7401a7b41648102)
Signed-off-by: Henning Schild <henning.schild@siemens.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Insert an explicit pass to fetch all blobs needed by Git LFS, during the
fetch() function. This avoids the default behavior of Git LFS to wait
until 'git checkout' to begin downloading the blobs pointed to by LFS records.
Network access is not allowed at that point in the recipe's lifecycle.
[YOCTO #14191]
(Bitbake rev: 0efac66043662e7a2027192f50e92e982db2ba1c)
Signed-off-by: Matt Hoosier <matt.hoosier@garmin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The usehead url parameter for git repositories causes bitbake to use
whatever commit the repository HEAD is pointing to if the repository
happens to have the name 'default'. This is the default name so in many
cases it works just fine, but if a different name is specified with the
url parameter 'name=newName' then it will fail to parse the recipe with
an error along the lines of:
ERROR: ExpansionError during parsing /path/to/my/recipe.bb
Traceback (most recent call last):
File "/path/to/poky/bitbake/lib/bb/fetch2/git.py", line 235, in Git.urldata_init:
> ud.setup_revisions(d)
File "/path/to/poky/bitbake/lib/bb/fetch2/__init__.py", line 1302, in FetchData.setup_revisions:
for name in self.names:
> self.revisions[name] = srcrev_internal_helper(self, d, name)
File "/path/to/poky/bitbake/lib/bb/fetch2/__init__.py", line 1167, in srcrev_internal_helper(name='newName'):
if srcrev == "AUTOINC":
> srcrev = ud.method.latest_revision(ud, d, name)
File "/path/to/poky/bitbake/lib/bb/fetch2/__init__.py", line 1562, in Git.latest_revision(name='newName'):
except KeyError:
> revs[key] = rev = self._latest_revision(ud, d, name)
return rev
File "/path/to/poky/bitbake/lib/bb/fetch2/git.py", line 650, in Git._latest_revision(name='newName'):
raise bb.fetch2.FetchError("Unable to resolve '%s' in upstream git repository in git ls-remote output for %s" % \
> (ud.unresolvedrev[name], ud.host+ud.path))
bb.data_smart.ExpansionError: Failure expanding variable SRCPV, expression was ${@bb.fetch2.get_srcrev(d)} which triggered exception FetchError: Fetcher failure: Unable to resolve 'master' in upstream git repository in git ls-remote output for /path/to/local/git/repo
Let's fix this by setting the unresolved rev of _all_ repository names
to 'HEAD' when the usehead url parameter is specified. Update the
currently failing test, test_local_gitfetch_usehead_withname, to now
expect success.
This change preserves existing behavior that allows usehead to be
overridden by a valid looking revision if one happens to be specified
instead of AUTOREV.
(Bitbake rev: 01e901c44ab0f496606b1d45c8953dc54970204c)
Signed-off-by: Joey Degges <jdegges@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This keeps coming up, document why we don't do this and the alternatives.
(Bitbake rev: aded964eed4ce5a725ed1ab477efabc86b1aa481)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This commit replaces the instances where escaped double quotes
are used to support SRC_URI url containing spaces with the more
pythonic shlex.quote().
(Bitbake rev: 4f9ba9c794de55bea0343267467bddea99844374)
Signed-off-by: Charlie Davies <charles.davies@whitetree.xyz>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Microsoft's TFS VCS system allows for spaces in a git repository url.
An example of a valid url is:
ssh://tfs-my-company.org:22/tfs/My Projects/FooBar
This commit adds support for such urls by implementing two changes.
Firstly, when bitbake makes a git command line call the url is
surrounded by quotes so that the url, regardless of spaces, is
treated as one argument. Secondly, additional parsing of various
filepath variables, which are based off of the url, are now
completed with any spaces in the url replaced with underscores.
(Bitbake rev: eb38b6f0935763f7ba19e5618f376fcae1dac41a)
Signed-off-by: Charlie Davies <charles.davies@whitetree.xyz>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It looks like we're about to see a lot of changes in branch names in repos. If
we have the prune option here, those old names are lost, the changes propagate
to our source mirrors and our old releases break.
We have the force option so any replaced references should be replaced, its only
orphaned branches which will now be preserved.
I believe this behaviour will cause us fewer problems given the changes that
look likely to happen.
(Bitbake rev: 820ab886e79eea516560c0c008e4cf059c6e11a3)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Function "contains_lfs" was only looking at the master branch when searching for LFS
content. LFS may be configured in specific branches only, so we need to use the
correct branch.
(Bitbake rev: 4fa67c2af830035a1ddedc14592ee25a15ebff22)
Signed-off-by: Mauro Queiros <maurofrqueiros@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The message "Repository %s has LFS content but it is not being fetched" was
being printed, even when Git-LFS was available and "lfs=1" was set. In those
situations, we want to fetch LFS content, so that message would not make sense.
(Bitbake rev: 45028dfda5a29a34ab408cb3f11d72ae17963340)
Signed-off-by: Mauro Queiros <maurofrqueiros@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Git-LFS objects were being fetched even when lfs=0 was not set.
This patch disables LFS smudging when lfs=0. That way, only the LFS pointers
are downloaded during checkout.
(Bitbake rev: 646d86df7de774255246a3d7051c308e43eb257d)
Signed-off-by: Mauro Queiros <maurofrqueiros@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
>From a SRCREV caching point of view, there is no reason to treat the
following upstreams as different:
SRC_URI = "git://github.com/file/file.git"
SRC_URI = "git://github.com//file/file.git"
(Bitbake rev: 425e21c14955dd38868c6e97637df3bbe0f89fac)
Signed-off-by: Chris Laplante <chris.laplante@agilent.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
amend the code to handle singleton comparisons properly so it only checks
if they only refer to the same object or not, and not bother
comparing the values.
(Bitbake rev: b809a6812aa15a8a9af97bc382cc4b19571e6bfc)
Signed-off-by: Frazer Clews <frazer.clews@codethink.co.uk>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When bitbake determines if a git clone needs updating, it only checks for the
needed srcrevs, not the revs listed in BB_GIT_SHALLOW_REVS, which will fail if
using shallow and the needed rev was added to the upstream git repo after a
previous fetch. Ensure that we also check for shallow revs.
[YOCTO #13586]
(Bitbake rev: 93b65e3c6ca64f644946953980595c44fbbcc748)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Refactor the git-lfs checking: this means both clearer code in the download()
function and allows unit testing to monkeypatch the functionality.
(Bitbake rev: 33cf9172ded50a869f7201ba463ab9ecc69b8252)
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Add a new 'lfs' option to the git fetcher so that we can optionally not fetch
git-lfs content, for repositories that contain LFS data that we don't
actually need for building.
By default lfs is set to 1, so if the repository has LFS content then git-lfs is
required. Setting lfs to 0 will mean that git-lfs won't be required to fetch,
and some files will be missing.
(Bitbake rev: be0b78ccfc5ede98041bc0545a15092494b12b26)
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
There are much better ways to handle this and most editors shouldn't need this
in modern times, drop the noise from the files. Its not consitently applied
anyway.
(Bitbake rev: 5e43070e3087d09aea2f459b033d035c5ef747d0)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
With the introduction of SPDX-License-Identifier headers, we don't need a ton
of header boilerplate in every file. Simplify the files and rely on the top
level for the full licence text.
(Bitbake rev: 695d84397b68cc003186e22f395caa378b06bc75)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This adds the SPDX-License-Identifier license headers to the majority of
our source files to make it clearer exactly which license files are under.
The bulk of the files are under GPL v2.0 with one found to be under V2.0
or later, some under MIT and some have dual license. There are some files
which are potentially harder to classify where we've imported upstream code
and those can be handled specifically in later commits.
The COPYING file is replaced with LICENSE.X files which contain the full
license texts.
(Bitbake rev: ff237c33337f4da2ca06c3a2c49699bc26608a6b)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Build will fail if repository has lfs contents in absense of git-lfs tool on host.
Build will pass if repository may or may not contains lfs content if host has git-lfs installed.
[YOCTO #13198]
(Bitbake rev: 3f2c2eb2f59707828bdcdd6414db837da8dc3b0e)
Signed-off-by: Naveen Saini <naveen.kumar.saini@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The fetch2/__init__.py checks whether "BB_FETCH_PREMIRRORONLY" == "1", but
fetch2/git.py and hg.py checks whether it is None, this makes it discontinuous,
and BB_FETCH_PREMIRRORONLY = "0" doens't work as expected in the later case,
so unify it to the previous one. (As BB_NO_NETWORK does).
And also use bb.utils.to_boolean() to make them consistent.
(Bitbake rev: 85a0d22835588e9ad8ec29d88a8115227e88420c)
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The localpath is a symlink to clonedir when it is cloned from a mirror, for
example:
$ bitbake systemtap-native -cfetch
$ ls downloads/git2
sourceware.org.git.systemtap.git -> /path/to/downloads/git2/mirror.path.git.sourceware.org.git.systemtap.git
mirror.path.git.sourceware.org.git.systemtap.git
There are both sourceware.org.git.systemtap.git and
mirror.path.git.sourceware.org.git.systemtap.git in DL_DIR/git2, the symlink
sourceware.org.git.systemtap.git is created by try_mirror_url(), but
do_cleanall" only removed the symlink, didn't remove the real dir
mirror.path.git.sourceware.org.git.systemtap.git, this may cause confusions,
for example, I assumed that do_cleanall removed everything, but it didn't, and
it would the re-used next time when do_fetch. This patch fixes the problem.
(Bitbake rev: 452e2200ad2c29dec3753f5f7a8cbc9183ec7dd8)
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
getVar() has been defaulting to expanding by default for
a long time (2016), thus remove the True option from
getVar() calls with a regex search and replace.
Search & replace made using the following command:
sed -e 's|\(d\.getVar \?\)( \?\([^,()]*\), \?True)|\1(\2)|g' \
-i $(git grep -E 'getVar ?\( ?([^,()]*), ?True\)' \
| cut -d':' -f1 \
| sort -u)
(Bitbake rev: 3bba0dbd524cf72176a765957adff544ae5c255a)
Signed-off-by: André Draszik <andre.draszik@jci.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Fix handling of escape characters in regexs and hence fix python
Deprecation warnings which will be problematic in python 3.8.
(Bitbake rev: c1fcc46e2498ddd41425d8756754f814d682aba3)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This code checks to see if shallow is either disabled or the tarball is
missing, but the else block tries to print the tarball filename, and
this attribute doesn't exist at all when shallow is disabled. Handle the
two cases separately to give sane errors for both cases without the
exception:
Exception: AttributeError: 'FetchData' object has no attribute 'fullshallow'
(Bitbake rev: bdbb558342ebb4e64384c9838d2485d9299d91a6)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Provide more information in the case the sources are not found in the
unpack step.
(Bitbake rev: 27a2214bf6f2e7c61bfc422a20959a55f7e0d25d)
Signed-off-by: Urs Fässler <urs.fassler@bbv.ch>
Signed-off-by: Pascal Bach <pascal.bach@siemens.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
A user friendly error is throw when neither the clonedir nor
fullshallow exist. Without the check, a difficult to interpret error
is throw from within the fetch command.
(Bitbake rev: 30cf2506007d25162f0805051212f54c39034ff3)
Signed-off-by: Urs Fässler <urs.fassler@bbv.ch>
Signed-off-by: Pascal Bach <pascal.bach@siemens.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The shallow_tarball check is never true due a check on the caller side.
The tarball check is not related to the code on the caller side.
(Bitbake rev: 086eddcf8c7520ff5c52ce2a11ca9bf5b5fe5d7e)
Signed-off-by: Urs Fässler <urs.fassler@bbv.ch>
Signed-off-by: Pascal Bach <pascal.bach@siemens.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
To improve the readability we extract the different scenarios of why
the clonedir needs an update.
(Bitbake rev: 9038e029f4f0ab413727de76c74248cbb3cdc9ea)
Signed-off-by: Urs Fässler <urs.fassler@bbv.ch>
Signed-off-by: Pascal Bach <pascal.bach@siemens.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Before this fix it is assumed that the removal of the
remote can only fail because there is not remote to remove. This
is a false assumption. Example error which would be ignored:
git -c core.fsyncobjectfiles=0 remote rm origin failed with exit code 1, output:
Note: A branch outside the refs/remotes/ hierarchy was not removed;
to delete it, use:
git branch -d master
error: could not lock config file config
error: Could not remove config section 'remote.origin'
Due to the masking of this error a stranger error will be
presented to the user, because this time we do not mask the
exception:
git -c core.fsyncobjectfiles=0 remote add --mirror=fetch origin https://github.com/ptsneves/tl-wn722.git failed with exit code 128, output:
fatal: remote origin already exists.
The most likely reason that the remote cannot be removed nor
modified is that the DL_DIR/git2 does not have permissions
compatible with the user running bitbake.
This commit fixes:
https://bugzilla.yoctoproject.org/show_bug.cgi?id=12728
(Bitbake rev: 9c86c582a10c9b23abad7d34b6cbf12f7086294d)
Signed-off-by: Paulo Neves <ptsneves@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Currently there is quite some variation between the fetchers in terms
of how they determine the subdirectory within DL_DIR and the base
fetch command to run. Some rely on variables being set externally
(e.g. from bitbake.conf in oe-core), some respect these external
variables but provide fallback defaults and some use only hardcoded
internal values. Try to unify the approach used across the various
fetchers.
(Bitbake rev: efd5e35af4b08501c67e8b30f30d9457f6fdf610)
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Updating a local git repo clone currently results in multiple calls
to self._contains_ref(), some of which appear to be redundant and can
be eliminated by minor tweaks to the logic in download().
Also drop redundant calls to os.path.exists(ud.clonedir) before
self.need_update(), since need_update() includes its own built-in
check for the existance of ud.clonedir.
(Bitbake rev: 61b0df5523afc8f805043f3adc9c106690e6f133)
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Move the code that existed in tests/fetch.py for determining the path to
'git-make-shallow' into the git module and reference it.
This ensures that 'git-make-shallow' is always available and the desired
version regardless of the path variable or whether git exposes the
command.
(Bitbake rev: 6b508ab8fd5aa796c1c00c970e81e5e93f84d35d)
Signed-off-by: Nathan Rossi <nathan@nathanrossi.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We've noticed issues on our infrastucture iterating over the many
tag/branch/head reference files that some git repositories may contain.
By issuing the pack-refs command, we move these all to a single file
which speeds up operations with the mirror repos in the downloads
directory in general.
(Bitbake rev: f8126aaf774186a6eaf0bd4067b89c074594886c)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Since 2017-08-17 (git version 2.14.1.473.g3ec7d702a) using deprecated
git branch parameter "--set-upstream" causes a fetcher error. Replace
it by "--set-upstream-to".
https://git.kernel.org/pub/scm/git/git.git/commit/?id=52668846ea2d41ffbd87cda7cb8e492dea9f2c4d
says, it's deprecated since 2012-08-30 so hopefully all still supported
host distributions have new enough git to support "--set-upstream-to".
ERROR: PACKAGE do_unpack: Fetcher failure: ...;
git -c core.fsyncobjectfiles=0 branch --set-upstream master origin/master failed with exit code 128, output:
fatal: the '--set-upstream' option is no longer supported. Please use '--track' or '--set-upstream-to' instead.
ERROR: PACKAGE do_unpack: Function failed: base_do_unpack
(Bitbake rev: 2ab50074c1a6c56a8a178755de108447d7b7acaf)
Signed-off-by: Andre Rosa <andre.rosa@lge.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
In certain cases, it's valuable to be able to exert more control over what
history is removed, beyond srcrev+depth. As one example, you can remove most
of the upstream kernel history from a kernel repository, keeping predominently
the non-publically-accessible content. If the repository is private, the
history in that repo couldn't be restored via `git fetch --unshallow`, but
upstream history could be.
Example usage:
# Remove only these revs, not at a particular depth
BB_GIT_SHALLOW_DEPTH_pn-linux-foo = "0"
BB_GIT_SHALLOW_REVS_pn-linux-foo = "v4.1"
(Bitbake rev: 97f856f0455d014ea34c28b1c25f09e13cdc851b)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
By default, all unused refs (branches & tags) are removed from the repository,
as shallow processing scales with the number of refs it has to process. Add
the ability to explicitly specify additional refs to keep. This is
particularly useful for recipes with custom checkout processes, or whose
git-based versioning requires a tag be available (i.e. for `git describe
--tags`). The new `BB_GIT_SHALLOW_EXTRA_REFS` variable is a space-separated
list of refs, fully specified, and support wildcards.
Example usages:
BB_GIT_SHALLOW_EXTRA_REFS = "refs/tags/v1.0"
BB_GIT_SHALLOW_EXTRA_REFS += "refs/heads/*"
(Bitbake rev: 1771934cd9f8b5847c6fcae0a906fb99d6b0db16)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Allow the user to explicitly adjust the depth for named urls/branches. The
un-suffixed BB_GIT_SHALLOW_DEPTH is used as the default.
Example usage:
BB_GIT_SHALLOW_DEPTH = "1"
BB_GIT_SHALLOW_DEPTH_doc = "0"
BB_GIT_SHALLOW_DEPTH_meta = "0"
(Bitbake rev: 9dfc517e5bcc6dd203a0ad685cc884676d2984c4)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This adds support to the git fetcher for fetching, using, and generating
mirror tarballs of shallow git repositories. The external git-make-shallow
script is used for shallow mirror tarball creation.
This implements support for shallow mirror tarballs, not shallow clones.
Supporting shallow clones directly is not really doable for us, as we'd need
to hardcode the depth between branch HEAD and the SRCREV, and that depth would
change as the branch is updated.
When BB_GIT_SHALLOW is enabled, we will always attempt to fetch a shallow
mirror tarball. If the shallow mirror tarball cannot be fetched, it will try
to fetch the full mirror tarball and use that. If a shallow tarball is to be
used, it will be unpacked directly at `do_unpack` time, rather than extracting
it to DL_DIR at `do_fetch` time and cloning from there, to keep things simple.
There's no value in keeping a shallow repository in DL_DIR, and dealing with
the state for when to convert the clonedir to/from shallow is not worthwhile.
To clarify when shallow is used vs a real repository, a current clone is
preferred to either tarball, a shallow tarball is preferred to an out of date
clone, and a missing clone will use either tarball (attempting the shallow one
first).
All referenced branches are truncated to SRCREV (that is, commits *after*
SRCREV but before HEAD are removed) to further shrink the repository. By
default, the shallow construction process removes all unused refs
(branches/tags) from the repository, other than those referenced by the URL.
Example usage:
BB_GIT_SHALLOW ?= "1"
# Keep only the top commit
BB_GIT_SHALLOW_DEPTH ?= "1"
# This defaults to enabled if both BB_GIT_SHALLOW and
# BB_GENERATE_MIRROR_TARBALLS are enabled
BB_GENERATE_SHALLOW_TARBALLS ?= "1"
(Bitbake rev: 5ed7d85fda7c671be10ec24d7981b87a7d0d3366)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Remove ud.mirrortarball in favor of ud.mirrortarballs. Each tarball will be
attempted, in order, and the first available will be used. This is needed for
git shallow mirror tarball support, as we want to be able to use either
a shallow or full mirror tarball.
(Bitbake rev: 02eebee6709e57b523862257f75929e64f16d6b0)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We call git ls-remote to get the latest revision from a git repository,
however by calling runfetchcmd() we can end up recursively running
git ls-remote a number of times with OE e.g. if ${SRCPV} is in PV, ${PV}
is in WORKDIR, and ${WORKDIR} is in PATH (as a result of recipe-specific
sysroots), our call to runfetchcmd() exports PATH so _lsremote() will
get called again - with the end result that we run git ls-remote 30
times in quick succession (!). Prevent that from happening by using a
guard variable and returning a dummy value if it's called recursively.
Fixes [YOCTO #11185].
(Bitbake rev: ff1ccd1db5d70b3fc9ad0d3e8f3d7b804c22bf36)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
FetchError isn't defined, use bb.fetch2.FetchError in this context.
(Bitbake rev: 945fa980e027753df2c21d84eb63dcaddb2caaee)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The touch of .done explicitly specifies the path, so there's no need for
workdir=, and "os.path.join('.')" is identical to just '.'.
(Bitbake rev: 955cbfdaa2400d15ec428b65848e6835c9f44860)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
list.index() isn't a particularly efficient operation, so keep track of our
position via enumerate() instead, which is more pythonic as well.
(Bitbake rev: dec6e90a4d27ee335e9c78aeebd277098fec94d1)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Cleanup some more usage of bb.data APIs in the fetchers.
(Bitbake rev: 9752fd1c10b8fcc819822fa6eabc2c1050fcc03b)
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
For spelling's sake, rename Python routine "setup_revisons" to
"setup_revisions."
(Bitbake rev: 4df59b027c02ef39d72476251ccd3fd62fc20bf6)
Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(Bitbake rev: 05f5421b2e44cd58c5912848de43d5884d070150)
Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
getVar() now defaults to expanding by default, thus remove the True
option from getVar() calls with a regex search and replace.
Search made with the following regex: getVar ?\(( ?[^,()]*), True\)
(Bitbake rev: 3b45c479de8640f92dd1d9f147b02e1eecfaadc8)
Signed-off-by: Joshua Lock <joshua.g.lock@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The old style bb.data.getVar/setVar API is obsolete. Most of bitbake
doesn't use it but there were some pieces that escaped conversion. This
patch fixes the remaining users mostly in the fetchers.
(Bitbake rev: ff7892fa808116acc1ac50effa023a4cb031a5fc)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Fix the methods in all fetchers so they don't change
the current working directory of the calling process, which
could lead to "changed cwd" warnings from bitbake.
(Bitbake rev: 6aa78bf3bd1f75728209e2d01faef31cb8887333)
Signed-off-by: Matt Madison <matt@madison.systems>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Introduce a new 'usehead' url parameter for git repositories. Specifying
usehead=1 causes bitbake to use whatever commit the repository HEAD is
pointing to. Usage of usehead=1 is only allowed for local git
repositories, i.e. it must always be accompanied with protocol=file url
parameter.
[YOCTO #9351]
(Bitbake rev: 2673fac5a9d06de937101e3fb2ddf1e60ff99abf)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Markus Lehtonen <markus.lehtonen@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Implement progress reporting support specifically for the fetchers. For
fetch tasks we don't necessarily know which fetcher will be used (we
might initially be fetching a git:// URI, but if we instead download a
mirror tarball we may fetch that over http using wget). These programs
also have different abilities as far as reporting progress goes (e.g.
wget gives us percentage complete and rate, git gives this some of the
time depending on what stage it's at). Additionally we filter out the
progress output before it makes it to the logs, in order to prevent the
logs filling up with junk.
At the moment this is only implemented for the wget and git fetchers
since they are the most commonly used (and svn doesn't seem to support
any kind of progress output, at least not without doing a relatively
expensive remote file listing first).
Line changes such as the ones you get in git's output as it progresses
don't make it to the log files, you only get the final state of the line
so the logs aren't filled with progress information that's useless after
the fact.
Part of the implementation for [YOCTO #5383].
(Bitbake rev: 4027649f422ee64b1c4e1ad8d48ac295050afbff)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Previously the code used to match a reference to its SHA-1 in
_latest_revision() used the Python "in" operator, which made it match
if the reference matched the beginning of an existing tag or
branch. This test, however, must be exact. I.e., either the reference
matches a tag or branch exactly, or it does not match at all.
(Bitbake rev: e5986c78a6108fd7578989c20efcbf0b81c97e03)
Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It was used for workaround git 1.7.9.2 which was released in 2012 which
should not be existed on nowadays host, so remove it to avoid
confusions.
(Bitbake rev: 6140d0cc9aecf1029ca16fed47071dfcc92f4269)
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This reverts commit 4193e99adce8e88f12ac88d7578ad39575f7e346.
It seems the underlying issue was caused by ":" in the url which isn't
supported. The patch was therefore incorrect.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This goes undetected most of the time, but when updating a repository,
if the ud.fullmirror file is not present, you end up getting an
exception instead of carrying on because the errno module is not
loaded (specifically "if exc.errno != errno.ENOENT").
(Bitbake rev: e6fca8480731ce817df9bee61438347a5e3d3017)
Signed-off-by: Kristian Amlie <kristian.amlie@mender.io>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Some servers, e.g. bitbucket.org can't cope with ssh:// as part of
the git url syntax. git itself is happy enough with this but you
get server side errors when using it.
This changes the git fetcher to use the more common ssh url format
which also means we need a : before the path.
Seems a shame to have to do this due to broken servers however
it should be safe enough since this other form is the one most people
use on the commandline so it should be safe enough.
[YOCTO #8864]
(Bitbake rev: 4193e99adce8e88f12ac88d7578ad39575f7e346)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Rename REGEX to UPSTREAM_CHECK_REGEX, REGEX_URI to UPSTREAM_CHECK_URI, and
GITTAGREGEX to UPSTREAM_CHECK_GITTAGREGEX to better reflect their purpose
and to reflect a common namespace.
(Bitbake rev: f0a9e783f9969573fd74edfa241ef14f18ac684e)
Signed-off-by: Alexander Kanavin <alexander.kanavin@linux.intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Currently this module is dereferencing errno.ENOENT but the python module "errno"
is not imported, which causes a crash when fetching from a git repository.
(Bitbake rev: 93e4c9bb2393b1074f5a01e7eaaac742a59d8086)
Signed-off-by: Romain Perier <romain.perier@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We were maintaining state in the form of ud.repochanged to determine whether
we need to write the tarball in the case where the tarball already exists, but
this state is maintained only in memory. If we need to update the git repo,
set ud.repochanged to True, and then are interrupted or killed, the tarball
will then be out of date.
Rather than maintaining this state, simply remove the out of date tarball when
we update the git repo, and it will be re-created with updated content in
build_mirror_data.
[YOCTO #6366]
(Bitbake rev: eaaa81393f181432c8586b17ade623f42c9fed2e)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When using an absolute file URI, there's no host, and the path starts with
'/', the dir under ${DL_DIR}/git2/ ends up starting with '.', so is hidden.
Remove any leading '.' to fix this.
(Bitbake rev: 8dce6964d56b36a77fb113f2ad496cc992a5ff36)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We need to know the revision associated with the upstream version in
SCM like git, this change broke return convention, oe-core recipeutils
get_recipe_upstream_version need to be updated.
tests/fetch.py: Updated git latest_versionstring test for comply new
convention.
[YOCTO #7605]
(Bitbake rev: fd175dc90024c503134c11cbd83e77d88c406ac8)
Signed-off-by: Aníbal Limón <anibal.limon@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Don't limit the tag search for only tags end with ^{}.
(Bitbake rev: 7006ab313766344cf33481228465082ed5977d28)
Signed-off-by: Aníbal Limón <anibal.limon@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
* Create a branch and named as upstream branch when checkout source
* Set the branch to track remote branch.
(Bitbake rev: 1ba20e4fe9c884515b200589fe379ad5eeda10bd)
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
In order to pass connection cache object to checkstatus function
add fetch parameter.
(Bitbake rev: fbb9c6f5538084e125b58118a86968908e6f895b)
Signed-off-by: Aníbal Limón <anibal.limon@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
gitpkgv_revision returns a sortable revision number that can be used
in the PKGV variable for example. To mimic meta-openembedded gitpkgv
behaviour to provide a sortable revision numner, one could set the
following:
PKGV = "1.0+${@bb.fetch2.get_srcrev(d, 'gitpkgv_revision')}"
This would yield a package version like "1.0+69+fb5eb80".
(Bitbake rev: 989c08f62aff7b707c25c692c23284f16506b7bc)
Signed-off-by: Mike Looijmans <mike.looijmans@topic.nl>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Pass proper repository url without arguments after a semicolon.
Executing checkuri on a rule with git repository in SRC_URI does
not report errors when working offline because wrong repository
url is passed to the ls-remote command. For example
"bitbake -c checkuri glibc" command executes:
"git -c core.fsyncobjectfiles=0 ls-remote git://sourceware.org/git/glibc.git;branch=release/2.21/master"
command in a shell subprocess to determine if url is valid.
Shell subprocess executes in fact 2 commands:
"git -c core.fsyncobjectfiles=0 ls-remote git://sourceware.org/git/glibc.git"
and
"branch=release/2.21/master"
First one returns 127 or 128 depending on error but second one
returns 0 because it is just env variable setup. Therefore we're not catching
connection error.
[YOCTO #7558]
(Bitbake rev: efa44d04137977f883db4a643b0f774e91514722)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The URL is not sent when _latest_revision generates and exception.
When performing the sanity checks it is not possible to know the URI that failed.
This add the URL when latest_revision generates an exception.
[YOCTO: #7592]
(Bitbake rev: 9f2115b07a55cb14e4a74dc6fbd3707c28a234d0)
Signed-off-by: Mariano Lopez <mariano.lopez@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
If you're interested in using the checked out repository for development
(e.g. in OE with devtool) then you ideally want the origin remote to
point to the repository it was fetched from, so just set that after
cloning.
(As part of this I did a minor refactor so we have one function to
generate the repository URL, which was already in two places.)
Fixes [YOCTO #7756].
(Bitbake rev: 80ecd1c54d4c748cee3a7ce0d64013a346e7671e)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
If the subpath parameter to the git fetcher ends with a trailing '/',
bb.utils.prunedir() will be called on '/'...
Fixes [YOCTO #7620].
(Bitbake rev: 380a3fb372c8b0a53dd7528562e6e7a222dc76ef)
Signed-off-by: Anders Darander <anders@chargestorm.se>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Currently the code ignores lightweight tags which has caused some user
complaints. We can't put the right search list in place easily since
the results don't come back in a good order, head happens to sort
before tags.
In the end I refactored the function so we get the complete list of
remotes and then we can filter it ourselves in the order we chose,
including checking for light weight tags, preferring the proper ones.
Hopefully this resolves the issues people have been seeing.
[YOCTO #6881]
(Bitbake rev: 07ad307065bb15a48f0015b9e4a643201abdc283)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Being able to generate a version string representing the most recent git commit
given git is useful, not least for the package reporting system.
This adds in a latest_versionstring method to the git fetcher
which allows users to query the latest version using ls-remote
and filtering the responses.
The patch also adds unittests for this function so that if
improvements are made, the original test urls can be used
to evaulate the those changes.
This is based on code from Irina Patru <irina.patru@intel.com>.
(Bitbake rev: f71c8c0354e87fed80bc845db6728e6e18ce9c4d)
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Aníbal Limón <anibal.limon@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This makes it possble to fetch Gerrit review references which are
normally stored under refs/changes.
Please disregard previous patch with the same topic.
(Bitbake rev: 268e9c0c6830e8e621c418f20c2ca12dc840e48b)
Signed-off-by: Fredrik Svensson <fredrik.svensson@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We only ever clone other repositories, if there were a problem such as power
failure, we'd blow away data and rebuild. As such we don't need fsync(). With
filesystems like ext*, the fsync pushes nearly all the data out to disk
which impacts all running processes.
We therefore set a configuration parameter to disable the fsync() calls.
Also fixup a case where basecmd wasn't being used for no good reason.
(Bitbake rev: 0a26abaf3a1e34d556c9375068dd17c879568d0f)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
There is other code which can want to run ls-remote style commands with
different parameters so split out the function.
(Bitbake rev: 13f1138f5504feee0ee8e8f3a0675d0bea490351)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We need to deference tags when trying to map them to commit IDs with
ls-remote. If we don't do this, a given commit might not show up
later in a specific branch. There appears to be no good reason not
to do this.
(Bitbake rev: 8ef24f4c834298348172b96ec0b855bf09552b09)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When specifying tags, they're searched for unanchored so foo/bar could
match:
refs/heads/abc/foo/bar
refs/heads/xyz/foo/bar
refs/heads/foo/bar
This change anchors the expressions so they are based against heads
or tags (or any other base level tree that has been created).
(Bitbake rev: df2e0972cd1db7abd5ec8b7cb295fb0c42e284a4)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
For rebased git tree, some commits doesn't exist in any branch, and such commits are
valid in tag, the change is useful for such case.
(Bitbake rev: f594cb9f5a18dd0ab2342f96ffc6dba697b35f65)
Signed-off-by: Zhenhua Luo <zhenhua.luo@freescale.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
There was a problem:
$ bitbake xf86-video-omapfb -cfetch && bitbake xf86-video-omapfb -ccleanall
The git2_git.pingu.fi.xf86-video-omapfb.tar.gz has been removed from the
DL_DIR, but the git2_git.pingu.fi.xf86-video-omapfb.tar.gz.done still exists,
this is because the "open(ud.donestamp, 'w').close()" in try_mirror_url() will
create the git2_git.xxx.tar.gz.done, but no one removes it (the clean() in
fetch2/__init__.py removes the DL_DIR/git2/pkg.done)
This only happens on the git fetcher AFAIK.
[YOCTO #5688]
(Bitbake rev: fb2dc84875eb477661f421b21bc404d4805ce379)
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The basecmd is initialized in urldata_init; there's no need redoing that
work.
(Bitbake rev: f8df6f746fb2e27f029a5449cee6c891b1f36f4f)
Signed-off-by: Olof Johansson <olof.johansson@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Currently the fetcher doesn't distinguish between names that the fetcher
needs to resolve verses branch names that the user specified.
This meant that if you specify a tag and a branch, the fetcher broke. This
separates the two so that the branch name is preserved and can be used in
appropriate places.
(Bitbake rev: e85f39fe9d1b224414b5da0780da514f75c5df92)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The fetcher made the rather bold assumption that if it fetched from the upstream,
the revisions were present and correct. These checks are fast and ensure that
really is the case. The avoids accidental network accessed and missing
branch configuration problems.
(Bitbake rev: a9112a102a89049cda597dad449e922c9e957a5d)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
