| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |
|
|
|
| |
Ensure all objects have been sent before moving them into the repository
and do not terminate pusher while receiver is still processing.
|
| |
|
|
|
|
| |
Retry network tasks up to two times by default.
Fixes #30
|
| |
|
|
|
| |
Move parent_start_listening() from __init__ to spawn() to support
respawning a job after shutdown.
|
| | |
|
| |
|
|
|
|
| |
Unexpected connection termination should not be considered a bug.
Fixes #51
|
| |
|
|
|
| |
If --on-error is specified to decide the failure action on the command line,
then dont interactively handle that decision.
|
| |
|
|
|
| |
Before we were assuming the user would specify a cwd which exists,
now we dont care and just create it if it's not there.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
The pusher already checks this and the check in the receiver does not
provide any additional guarantees as it is prone to race conditions.
This prevents a push error in case two clients push an artifact with the
same key around the same time.
Fixes #52
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Make the main BuildStream process the foreground process again when the
interactive child exits. Otherwise the next read() on stdin will trigger
SIGTTIN and stop the process. This is required because the sandboxed
process does not have permission to do this on its own (running in
separate PID namespace).
dash still prints an error because it fails to restore the foreground
process, however, this is harmless. bash doesn't print an error in this
case, but the behavior is otherwise identical.
Fixes #41
|
| |
|
|
| |
Fixes #49
|
| |
|
|
|
| |
Add an extra argument to the function to know which elements
were already resolved.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously, the first time you configured an artifact cache, you would
get to the end of your first build and then BuildStream would exit
because of some stupid mistake like you got the address slightly wrong
or you forgot to add the host keys of the remote artifact cache to
`~/.ssh/known_hosts`.
To avoid surprises, if there's an artifacts push-url configured we now
try to connect to it as a preflight check so that issues are raised
early.
On success, you will see something like this:
[--:--:--][90904fe4][ main:gnu-toolchain/stage2.bst ] START Checking connectivity to remote artifact cache
[00:00:00][90904fe4][ main:gnu-toolchain/stage2.bst ] SUCCESS Connectivity OK
On failure, it looks like this:
[--:--:--][90904fe4][ main:gnu-toolchain/stage2.bst ] START Checking connectivity to remote artifact cache
[00:00:03][90904fe4][ main:gnu-toolchain/stage2.bst ] FAILURE BuildStream will be unable to push artifacts to the shared cache: ssh: connect to host ostree.baserock.org port 2220: Connection timed out
As a bonus, for some reason this check causes SSH to ask about unknown
host keys rather than just failing, so you may now see messages like
this if the host keys are unknown rather than an error:
The authenticity of host '[ostree.baserock.org]:22200 ([185.43.218.170]:22200)' can't be established.
ECDSA key fingerprint is SHA256:mB+MNfYREOdRfp2FG6dceOlguE/Skd4QwnS0tvCPcnI.
ECDSA key fingerprint is MD5:8f:fa:ab:90:19:31:f9:f7:f1:d4:e5:f0:a2:be:56:71.
Are you sure you want to continue connecting (yes/no)?
|
| |
|
|
|
|
|
| |
The full ref list can easily exceed the maximum message size. Limit list
to refs being pushed.
Fixes #47
|
| |
|
|
|
| |
If push-url is specified, it must point to the same repository as
pull-url as the summary file is used for pull and push operations.
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
They are no longer needed.
|
| |
|
|
| |
_set_built() is being removed.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
Build planning uses list of artifacts in remote artifact cache. Pull
failures cannot be ignored.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
OSTree fetch operations in subprocesses hang if OSTree fetch operations
have been used in the main process.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Suggest the user to mount /etc/passwd inside the container to fix `bzr`
command.
The command `bzr` fails if the uid of the user using it doesn't exist,
giving, for example, the following back trace when running `bzr init`:
File "/usr/lib64/python2.7/site-packages/bzrlib/lockdir.py", line 238, in _attempt_lock
tmpname = self._create_pending_dir()
File "/usr/lib64/python2.7/site-packages/bzrlib/lockdir.py", line 335, in _create_pending_dir
info = LockHeldInfo.for_this_process(self.extra_holder_info)
File "/usr/lib64/python2.7/site-packages/bzrlib/lockdir.py", line 779, in for_this_process
user=get_username_for_lock_info(),
File "/usr/lib64/python2.7/site-packages/bzrlib/lockdir.py", line 863, in get_username_for_lock_info
return osutils.getuser_unicode()
File "/usr/lib64/python2.7/site-packages/bzrlib/osutils.py", line 356, in _posix_getuser_unicode
name = getpass.getuser()
File "/usr/lib64/python2.7/getpass.py", line 158, in getuser
return pwd.getpwuid(os.getuid())[0]
KeyError: 'getpwuid(): uid not found: 1000'
To fix this there were 2 possible solutions:
- To run everything as rooat (or as another user) in the container. We
will not need to set the uid to use, making the instructions simpler.
The main problem is that the files stored in the host will probably
have a different owner, and the user will have problems with file
permissions.
- To ensure the uid of the current user also exists inside the
container.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
It takes public data from the input element and uses that to generate a
debian package.
|
| |
|
|
|
| |
It produces an artifact, plus write public data to provide enough
information to build a debian package out of the artifact.
|
| |
|
|
|
|
|
|
|
|
| |
The `list` function was already named `list_` to avoid clashing
with the builtin python type, the `open` command was causing log
viewing to fail because we try to call the `open` builtin python
function but it's shadowed.
So just call all of the workspace functions with a `workspace_`
prefix (but keep the same names on the command line).
|
| | |
|
| | |
|
| | |
|
