git archive capability works?

[bestander]

Do you want to request a feature or report a bug?

bug?

What is the current behavior?

Git archive capability checks that URL matches USER@HOSTNAME:PATHNAME
https://github.com/yarnpkg/yarn/blob/master/src/util/git.js#L57

Is it something like ssh://git@bitbucket.org/bestander/test-repo.git:/?
The problem is that :PATHNAME does not work with git-archive feature.
Even if the regex matches it will fail git-archive command check and the git.js will fallback to cloning.

Removing :PATHNAME from the regex could be a solution but I am not sure doing only this fixes the issue because supportsArchiveCache stores archive capability per hostname but archive capability also depends on access rights and protocol.

This feature needs a fix + tests.

Please mention your node.js, yarn and operating system version.

Yarn 0.22.0

[BYK]

@bestander I think this is now fixed via this method:

yarn/src/util/git.js

Lines 128 to 146 in ba8d0db

	static async hasArchiveCapability(ref: GitUrl): Promise<boolean> {
	const hostname = ref.hostname;
	if (ref.protocol !== 'ssh:' || hostname == null) {
	return false;
	}
	if (hostname in supportsArchiveCache) {
	return supportsArchiveCache[hostname];
	}
	try {
	await spawnGit(['archive', `--remote=${ref.repository}`, 'HEAD', Date.now() + '']);
	throw new Error();
	} catch (err) {
	handleSpawnError(err);
	const supports = err.message.indexOf('did not match any files') >= 0;
	return (supportsArchiveCache[hostname] = supports);
	}
	}

Closing the issue but LMK if you think this still needs more work or attention.

[Et7f3]

cross post of a comment on a commit (might not have visibility and is code from 9 years ago)
Sorry for necro-mancing but can't find info on google/git manual. If I understand this functionality avoid repacking each time ? but how it know which tarball to give back because it use Date.now(). I would have expected a hash or something more stable like tag/branch.

yarn/src/util/git.js

Line 152 in 158d96d

await spawnGit(['archive', `--remote=${ref.repository}`, 'HEAD', Date.now() + '']);

Do you know a service that support it ?

gitlab.com support it (and probably all gitlab instance).
I haven't tried gitolite but does it support this feature ?
local git ?

EDIT: after testing command on gitlab.com and github.com I finally understood.
We query a file that have very low chance to exist (Date.now()). The error message for github.com was:
fatal: operation not supported by protocol and the message in gitlab.com contains: fatal: operation not supported by protocol the string that is matched. Since it have a very low probability to exist it helps to discover if the functionality is available. Then we can use this capability to get only the file we want instead of full clone.

Very clever.