Respect `compression_format` from `containers.conf` in push, build `--cache-to`, and commit by Honny1 · Pull Request #6757 · containers/buildah

Honny1 · 2026-04-01T16:33:20Z

buildah push now falls back to compression_format and compression_level from containers.conf when --compression-format is not set
buildah build --cache-to passes CompressionFormat/ForceCompressionFormat to cache push, so cached layers use the configured compression instead of defaulting to gzip
buildah build gains --cache-compression-format, --cache-compression-level, and --cache-force-compression flags for controlling cache layer compression independently
buildah commit gains --compression-format, --compression-level, and --force-compression flags with containers.conf fallback
Fix commit.go switch to set DirForceCompress for all compression algorithms, not just gzip
Fix push.go blobcache to treat any non-uncompressed format as requiring compression
Add conflicting flag validation for --disable-compression with --compression-format/--force-compression

Fixes: #6660
Fixes: #6072

Depends on: containers/container-libs#731

What type of PR is this?

/kind api-change

/kind bug

/kind cleanup

/kind deprecation
/kind design
/kind documentation
/kind failing-test
/kind feature
/kind flake
/kind other

What this PR does / why we need it:

How to verify it

Which issue(s) this PR fixes:

Special notes for your reviewer:

Does this PR introduce a user-facing change?

`push`, build`, `commit`: respect `compression_format` from `containers.conf`

`buildah push`, `buildah build --cache-to`, and `buildah commit` now honor the `compression_format` and `compression_level` settings from `containers.conf`. Previously, `--cache-to` ignored these settings and always used `gzip`, causing cache layers to lose the configured compression. `buildah build` also gains new `--cache-compression-format`, `--cache-compression-level`, and `--cache-force-compression` flags for controlling cache layer compression independently of the final image. `buildah commit` also gains new `--compression-format`, `--compression-level`, and `--force-compression` flags.

Honny1 · 2026-04-01T16:45:41Z

While fixing the build cache, I decided to also check the pull and commit where compression is used.

packit-as-a-service · 2026-04-17T09:20:29Z

Ephemeral COPR build failed. @containers/packit-build please check.

Honny1 · 2026-04-17T10:17:22Z

/packit rebuild-failed

Honny1 · 2026-04-17T13:41:28Z

PTAL @containers/buildah-maintainers

simonbrauner · 2026-04-20T13:49:36Z

+	if defaultContainerConfig.Engine.CompressionLevel != nil {
+		compressionLevel = defaultContainerConfig.Engine.CompressionLevel
+	}


Suggested change

if defaultContainerConfig.Engine.CompressionLevel != nil {

compressionLevel = defaultContainerConfig.Engine.CompressionLevel

}

compressionLevel = defaultContainerConfig.Engine.CompressionLevel

Definitely non-blocking, it just caught my eye, but the if is redundant, right? compressionLevel would be nil on declaration, so assigning nil to it again would not be a problem.

simonbrauner · 2026-04-20T13:53:36Z

LGTM

nalind

What's the intent behind not adding to "build" the flags that are being added to "push" and "commit"?

nalind · 2026-04-20T21:19:41Z

+		if err != nil {
+			return err
+		}
+		options.CompressionFormat = &algo


Was the bit that checks for c.Flag("compression-format").Changed above intended to be merged into this block, to parallel the similar logic in the next block?

Ah, that is much cleaner. I will fix that.

nalind · 2026-04-20T21:33:53Z

 		if err != nil {
 			return err
 		}
 		options.CompressionFormat = &algo


Was the bit that checks for c.Flag("compression-format").Changed above intended to be merged into this block, to parallel the similar logic in the next block?

Honny1 · 2026-04-24T08:13:02Z

What's the intent behind not adding to "build" the flags that are being added to "push" and "commit"?

I forgot to do that. Good catch.

Honny1 · 2026-04-24T13:01:08Z

@nalind I addressed your comments.

Honny1 · 2026-04-27T16:17:55Z

PTAL @containers/buildah-maintainers

Honny1 · 2026-04-27T19:01:15Z

PTAL @containers/podman-maintainers

giuseppe

LGTM

nalind · 2026-04-27T21:55:00Z

+    $cid \
+    dir:$destdir
+
+  run cat $destdir/manifest.json


This should check "$status", just in case the "dir:" format changes, however unlikely that is.

nalind · 2026-04-27T22:01:43Z

 	fs.StringVarP(&flags.CWOptions, "cw", "", "", "confidential workload `options`")
+	fs.StringVar(&flags.CacheCompressionFormat, "cache-compression-format", "", "compression format to use for cache layers")
+	fs.IntVar(&flags.CacheCompressionLevel, "cache-compression-level", 0, "compression level to use for cache layers")
+	fs.BoolVar(&flags.CacheForceCompressionFormat, "cache-force-compression", false, "use the specified compression algorithm for cache layers if the destination contains a differently-compressed variant already")


build can write the final image to non-local destinations, as commit does. Could these flags also affect those cases?

These flags are intentionally cache-only (--cache-to). The final image commit already respects the compression_format in containers.conf, and buildah push has its own --compression-format flag for the actual push. Should I clarify this in the docs, or implement flags that affect the entire build?

Adding CLI flags to "commit" for overriding the defaults from the configuration file when committing the image, and having flags for "build" that do the same thing for cache images, but no equivalent for "build" for the final image just looks like an oversight. Most options for "commit" should also be available for "build".

I have added support for --compression-format, --compression-level, and --force-compression to apply to the final image during buildah build, not just cache layers. These take effect when the output is a non-local destination (registry, dir:, oci-archive:, etc.).

Honny1 · 2026-04-29T14:34:38Z

PTAL @nalind

Honny1 · 2026-05-04T14:29:46Z

PTAL @containers/buildah-maintainers @containers/podman-maintainers @mheon

Honny1 · 2026-05-13T14:21:19Z

PTAL @containers/buildah-maintainers @containers/podman-maintainers @nalind

nalind

Some questions about the tests that verify compression on --cache-to targets.

Honny1 · 2026-05-15T11:15:26Z

PTAL @containers/buildah-maintainers @containers/podman-maintainers @nalind

nalind

Some questions about the tests.

Honny1 · 2026-05-19T13:09:44Z

PTAL @containers/buildah-maintainers @containers/podman-maintainers @nalind

nalind

Some nits in the tests, but LGTM.

Signed-off-by: Jan Rodák <hony.com@seznam.cz>

Fixes: containers#6660 Fixes: containers#6072 Signed-off-by: Jan Rodák <hony.com@seznam.cz>

Signed-off-by: Jan Rodák <hony.com@seznam.cz>

Honny1 · 2026-05-20T12:13:48Z

PTAL @containers/buildah-maintainers @containers/podman-maintainers @nalind

nalind

LGTM, thanks!

Honny1 · 2026-05-20T14:00:56Z

PTAL @containers/buildah-maintainers @containers/podman-maintainers for final review and merge.

mtrmac · 2026-05-20T15:17:23Z

-func NewBlobCache(ref types.ImageReference, directory string, compress types.LayerCompression) (BlobCache, error) {
-	return imageBlobCache.NewBlobCache(ref, directory, compress)
+// The optional Option values can further refine behavior.
+func NewBlobCache(ref types.ImageReference, directory string, compress types.LayerCompression, opts ...Option) (BlobCache, error) {


(An absolutely non-blocking drive-by, by no means a review:) Maybe the Buildah callers could call the c/image package directly, without extending the API here.

These were left here to avoid breaking callers when the package's guts were moved.

Yes, I just meant that that doesn’t require adding more compatibility wrappers. The existing pkg/blobcache could remain as is for compatibility, and callers within Buildah could call the c/image one with the new options.

Anyway this already merged.

TomSweeneyRedHat · 2026-05-20T15:24:32Z

+	RewriteTimestamp       bool
+	CreatedAnnotation      bool
+	SourcePolicyFile       string
+	TransientRunMounts     []string


Not to fix now, but I will note that I hate seeing long structs like this that aren't in alpha order. It makes debugging later down the line harder IMHO>

TomSweeneyRedHat · 2026-05-20T15:26:24Z

LGTM

TomSweeneyRedHat · 2026-05-20T15:26:31Z

/lgtm

Honny1 closed this Apr 1, 2026

Honny1 force-pushed the teach-Zstd branch from 47e18c7 to 3f93dad Compare April 1, 2026 17:01

Honny1 reopened this Apr 1, 2026

Honny1 force-pushed the teach-Zstd branch 7 times, most recently from 4ce00cc to b66921a Compare April 17, 2026 09:14

Honny1 marked this pull request as ready for review April 17, 2026 10:11

dosubot Bot added the size:L This PR changes 100-499 lines, ignoring generated files. label Apr 17, 2026

simonbrauner approved these changes Apr 20, 2026

View reviewed changes

nalind reviewed Apr 20, 2026

View reviewed changes

Honny1 force-pushed the teach-Zstd branch from b66921a to eeb1767 Compare April 24, 2026 12:52

dosubot Bot added size:XL This PR changes 500-999 lines, ignoring generated files. and removed size:L This PR changes 100-499 lines, ignoring generated files. labels Apr 24, 2026

Honny1 requested a review from nalind April 24, 2026 12:57

Honny1 force-pushed the teach-Zstd branch from eeb1767 to 7bb49e3 Compare April 24, 2026 13:00

giuseppe approved these changes Apr 27, 2026

View reviewed changes

nalind requested changes Apr 27, 2026

View reviewed changes

Honny1 force-pushed the teach-Zstd branch from 7bb49e3 to 55c0a5c Compare April 28, 2026 10:08

Honny1 requested a review from nalind April 28, 2026 11:18

Honny1 force-pushed the teach-Zstd branch from 55c0a5c to de2045a Compare May 13, 2026 14:18

nalind requested changes May 13, 2026

View reviewed changes

Comment thread imagebuildah/stage_executor.go

Comment thread tests/bud.bats

Comment thread tests/bud.bats Outdated

Comment thread tests/bud.bats Outdated

Honny1 force-pushed the teach-Zstd branch 3 times, most recently from 39d0f40 to 94fb09a Compare May 15, 2026 09:46

nalind reviewed May 15, 2026

View reviewed changes

Comment thread tests/bud.bats

Comment thread tests/bud.bats

Comment thread tests/bud.bats Outdated

Comment thread tests/bud.bats Outdated

Comment thread tests/bud.bats Outdated

Comment thread tests/bud.bats Outdated

Honny1 force-pushed the teach-Zstd branch from 94fb09a to 0abf665 Compare May 19, 2026 08:37

nalind approved these changes May 19, 2026

View reviewed changes

Comment thread tests/commit.bats Outdated

Comment thread tests/push.bats

Comment thread tests/push.bats

Comment thread tests/bud.bats

Comment thread tests/bud.bats

Honny1 added 5 commits May 20, 2026 14:07

push: respect compression_format from containers.conf

ef4b7c1

Signed-off-by: Jan Rodák <hony.com@seznam.cz>

build: pass compression_format to --cache-to push

207c045

Fixes: containers#6660 Fixes: containers#6072 Signed-off-by: Jan Rodák <hony.com@seznam.cz>

commit: add --compression-format flag with containers.conf fallback

e6773b3

Signed-off-by: Jan Rodák <hony.com@seznam.cz>

build: add --compression-format flags for build

149909e

Signed-off-by: Jan Rodák <hony.com@seznam.cz>

docs: document compression flags for commit, push and build

bf9d66c

Signed-off-by: Jan Rodák <hony.com@seznam.cz>

Honny1 force-pushed the teach-Zstd branch from 0abf665 to bf9d66c Compare May 20, 2026 12:12

nalind approved these changes May 20, 2026

View reviewed changes

mtrmac reviewed May 20, 2026

View reviewed changes

TomSweeneyRedHat reviewed May 20, 2026

View reviewed changes

TomSweeneyRedHat merged commit 622a55a into containers:main May 20, 2026
40 checks passed

Conversation

Honny1 commented Apr 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What type of PR is this?

What this PR does / why we need it:

How to verify it

Which issue(s) this PR fixes:

Special notes for your reviewer:

Does this PR introduce a user-facing change?

Uh oh!

Honny1 commented Apr 1, 2026

Uh oh!

packit-as-a-service Bot commented Apr 17, 2026

Uh oh!

Honny1 commented Apr 17, 2026

Uh oh!

Honny1 commented Apr 17, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

simonbrauner commented Apr 20, 2026

Uh oh!

nalind left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Honny1 commented Apr 24, 2026

Uh oh!

Honny1 commented Apr 24, 2026

Uh oh!

Honny1 commented Apr 27, 2026

Uh oh!

Honny1 commented Apr 27, 2026

Uh oh!

giuseppe left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Honny1 commented Apr 29, 2026

Uh oh!

Honny1 commented May 4, 2026

Uh oh!

Honny1 commented May 13, 2026

Uh oh!

nalind left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Honny1 commented May 15, 2026

Uh oh!

nalind left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Honny1 commented May 19, 2026

Uh oh!

nalind left a comment

Honny1 commented Apr 1, 2026 •

edited

Loading