| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
| |
Fail instead of downloading a 404/403/500 whatever message.
Bug: https://bugs.gentoo.org/934308
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
| |
who am I kidding, the bootstrap and the mirror retaining is all
hardwired to bz2, and other compression formats don't really make a big
enough dent to warrant lots of changes to allow picking a different
compresion format
disable all but bz2, that's the only thing we use afterall
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
| |
lzip (lzma) compresses better than xz (lzma2) but xz can read/decompress
lzip archives, so drop the xz archive in favour of lzip
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
|
|
|
| |
gzip compresses to around 75MB which is much higher than bzip2's 57MB
The gzip snapshot was historically never used ever by the bootstrap
script, it always used and hardcoded bz2. So don't bother creating a
gzip variant.
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
Using -9 we end up at 58MB, which is more than gzip2's 57MB. So up the
level to -19 (the max, basically like how we configure the other
compressors too) to get 52MB, which is still much more than xz (48MB)
and lzip (47MB).
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
| |
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
| |
lzip seems to reach a better compression ratio than xz, let's see if we
can use it
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
| |
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
| |
Prepare for Zstandard becoming a standard, replacing another compressor.
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
| |
Disable false positives, unfortunately including the most helpful check
SC2086.
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
| |
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
| |
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
| |
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
| |
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
| |
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
| |
don't overwrite existing snapshots, only add so as to avoid changing
generated snapshots
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In bug #913415 a bunch of confusing and wrong things came out, and in an
attempt to fix this, the historical "prefix-overlay" fetched from dev
was replaced with ordinary "portage" from rsync snapshots dir. However,
this relies on the prefix mirror to provide the file, which can easily
be overridden by the user.
So retrieve the snapshot in use and retain it on the rsync slaves to
make this in the very least less confusing for those following the
locations, and ensure it works for people with custom mirror setups.
Bug: https://bugs.gentoo.org/913415
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
| |
Closes: https://bugs.gentoo.org/911543
Signed-off-by: Benda Xu <heroxbd@gentoo.org>
|
|
|
|
|
|
|
| |
use git reset --hard HEAD to bring the tree out of any conflict or
pollution state
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The action is failing at head, which makes it less than useful for
evaluating new changes.
e.g.
Error: ./scripts/rsync-generation/refresh-mirror.sh:44:2: error: Shells disambiguate (( differently or not at all. For subshell, add spaces around ( . For ((, fix parsing errors. [SC1105]
Error: ./scripts/rsync-generation/refresh-mirror.sh:44:3: error: Shells disambiguate (( differently or not at all. For subshell, add spaces around ( . For ((, fix parsing errors. [SC1105]
Error: ./scripts/bootstrap-prefix.sh:5:20: error: SIGKILL/SIGSTOP can not be trapped. [SC2173]
Error: ./scripts/bootstrap-prefix.sh:1669:19: error: Double quote array expansions to avoid re-splitting elements. [SC2068]
Error: ./scripts/bootstrap-prefix.sh:1964:19: error: Double quote array expansions to avoid re-splitting elements. [SC2068]
Error: ./scripts/bootstrap-prefix.sh:2286:19: error: Double quote array expansions to avoid re-splitting elements. [SC2068]
Error: ./scripts/bootstrap-prefix.sh:3194:11: error: -e doesn't work with globs. Use a for loop. [SC2144]
Signed-off-by: Peter Foley <pefoley2@pefoley.com>
Closes: https://github.com/gentoo/prefix/pull/19
Signed-off-by: Sam James <sam@gentoo.org>
|
|
|
|
|
|
|
| |
Somehow the git trees can get dirty, ensure they are completely clean
each time.
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
| |
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
|
|
|
| |
Now hashgen is integrated in portage-utils as qmanifest, switch to using
it.
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
|
|
|
| |
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
|
| |
|
|
|
|
| |
(filesdir) in apply_git_mtimes
|
|
|
|
| |
package removal
|
|
|
|
| |
hashgen will live on in github.com/grobian/hashgen
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
a logic rewrite was necessary to be able to properly generate the right
type of Manifest for the level at hand
|
| |
|
| |
|
|
|
|
|
| |
Tests were designed to expose problems, so it's kind of nasty when they
are included in the tree. Not a good idea.
|
|
|
|
|
|
| |
To get reproducible outputs, ensure we sort the directory listings, such
that the order and thus optional compression ratio is the same. This is
in particular nice when multiple generators are in use.
|
| |
|
|
|
|
|
|
| |
If we can't stat the file we're writing hashes for, we're not going to
produce anything useful and something up in the stack went wrong
already.
|
|
|
|
|
| |
relying on times stored in git is never going to work well, so pick a
time that we reasonably sync between multiple generators
|
| |
|
|
|
|
|
| |
save up subdirs to handle so we can nicely loop over them in a for-loop
which OpenMP can cheaply parallelise for us (for free)
|
| |
|
|
|
|
|
| |
Still not very happy about this, need a clearer way to deal with issues
to make a readable conclusion.
|
|
|
|
|
| |
also remove stray directories that for some reason where left behind, as
they cause verification problems
|
|
|
|
| |
This variant can verify the gx86 tree, or so it seems.
|
|
|
|
| |
missing digests
|
|
|
|
|
|
| |
When using Git, we use thin-manifests, but on the rsync tree, we
transformed the manifests into fat ones (hashgen), so we need to ensure
Portage knows that and performs its checks.
|
|
|
|
|
| |
AuthorDate can be anything, as I just found out, so don't rely on that.
Use CommitDate instead, which is hopefully closer to reality.
|