Age | Commit message (Collapse) | Author |
|
Exclude ZFS snapshot artifacts (`.poudriere-snap-*`) from the final
image.
|
|
Use `zpool create -t` to set the in-core pool name to a temporary,
random pool name, while the on-disk name will be the name specified as
the pool name in `ZFS_POOL_NAME`, by default `zroot` to match what is
currently used on release images.
Keep the check for a possible name collision on the randomly-created
pool name, to err on the side of caution.
Inspired by: https://reviews.freebsd.org/D34426
|
|
|
|
showconfig suggests 'make config' which can be confusing.
Fixes #992
|
|
|
|
-o port_dbdir will expand to $POUDRIERE_ETC/$port_dbdir. The
-jpz flags will be used for the make.conf, poudriere.conf, and
ports tree.
Fixes #111
Fixes #583
Fixes #176
|
|
|
|
Issue #985
|
|
Issue #985
|
|
Issue #985
|
|
Issue #985
|
|
Some older FreeBSD do not support passing this value over NFSv4
nor over NFSv3. This fixes wildly wrong durations showing in
some logs and statuses.
|
|
image_zfs: Fix setting boot partition size
|
|
|
|
Allow to disable grepping logs for failure reason
|
|
`zpool list -Ho ${ZFS_POOL_NAME}` exits with `0` if the name exists.
This prevents overwriting the existing pool.
|
|
|
|
|
|
- Auto-sized aligned columns using display_output()
- Add tmpfs space usage column
- Add cpu% and mem% (from ps) columns
|
|
|
|
|
|
|
|
|
|
|
|
|
|
- TMPFS_BLACKLIST should contain a list of package globs
- TMPFS_BLACKLIST_TMPDIR should contain a host directory prefix where
temporary directories can be created, outside tmpfs of course, to be
used as the WRKDIR for packages in TMPFS_BLACKLIST.
Fixes #888
|
|
If a SIGINT comes in while updating the relpaths it is possible
to end up in cleanup traps where the values do not match
$PWD.
|
|
|
|
|
|
|
|
|
|
Fixes: 80ec59d8e523fc116f80a8555a5d38bb60f09510
|
|
Fixes: 80ec59d8e523fc116f80a8555a5d38bb60f09510
|
|
|
|
This reverts commit 6d3fe23c9e43b9469536f330ede5c0015fd9c89f.
This is too racy with clean_pool() to not hide the errors from find.
|
|
Otherwise it may taint some lookups. Not sure how this has been
like this for so long.
|
|
|
|
|
|
This takes a lot of time for no usable result in most cases
|
|
Adds the ability to create a ZFS disk image in the same style as bsdinstall(8).
The default is to create a full GPT disk image, but is also supports just the raw pool.
Also adds `zfs send` streams as an output format, includes support for sending the
entire pool, just the boot environment (for upgrades), or both.
Supports specifying both zfs+send targets (+full and +be) in a single run.
Extends the existing -i support, to be able to modify existing images.
Adds the -R flag to allow user to specify ZFS send feature flags
Sponsored by: Modirum MDPay
Sponsored by: Klara Inc.
|
|
Replace `$(echo ${MEDIATYPE} | cut -d '+' -f 1)` with shell builtins
|
|
If the prebuild script sets SKIP_PREPARE then we don't run
${MAINMEDIATYPE}_prepare ${SUBMEDIATYPE}
Allowing users to replace the built-in _prepare for the image type
Cleans up the old 'skip' method that did nothing after b768a3692998
|
|
This ensures the hostname doesn't get set multiple times
|
|
missed in e71815b2a9b59fc569dbb04d971c51197d2ab550
|
|
This probably _never_ worked so add a low timeout too.
Fixes: 855b17c59b7
|
|
|
|
Packages are deleted if there is no bootstrap because we cannot read
them in delete_old_pkgs() and sanity_check_pkgs(). But we may have
fetched one that can work during download_from_repo().
This effectively reverts c25726ffc8033
|
|
I cannot get the logic here right. There are many complicated cases.
It is simpler if we just keep the bootstrap for metadata handling
and then build pkg on our own. This allows modifications to the
pkg port as well without needing more complexity in download_from_repo()
|
|
|
|
|