`constraint.single()` isn't preserving `version.raw`

[jhheider]

#196 (comment)

add console.info(pkg.version.raw) to brewkit/libexec/resolve.ts at line 17:

% ./libexec/resolve.ts zlib.net   
zlib.net=1.3.0
1.3

add console.info(version.raw) to brewkit/libexec/query.ts at line 44:

% ./libexec/query.ts zlib.net=1.3.0 --src

1.3.0

this was wreaking havoc with /pull/196 since it was testing generated distribution URLs (with HTTP HEAD requests), so calls to getDistributable() were failing. technically, they were always generating incorrect (but unused) URLs previously, and fetch.ts was using usePantry().resolve() which worked correctly. but this seems like a bug. right?

[mxcl]

well brewkit adding .raw is a hack to be fair. libpkgx doesn't support it. brewkit adds it to the object itself and it works because of coincidence.

edit: oh ok, no, raw is proper in this lib, I forgot and remembered something else.

[jhheider]

this was deep cut

[mxcl]

I need an example of it failing. This test passes:

assertEquals(new semver.Range("=1.0").single()!.raw, "1.0")

[mxcl]

k the bug is in getVersions. in brewkit.

[mxcl]

I'm not sure if this is a bug. The code gets the version from the source (eg github), the source dictates the raw version.

I need more details on why this is a bug in your opinion.

[isworking]

Take catb.org/wumpus and zlib.net as examples. The version it gets from the source is raw, which doesn't have .0 (in minor) at end. But when trying to fetch it, the .0 is appended (probably the existing version object is stringified?). It's breaking the builds and tests for them, at least. Idk if it's a bug in BrewKit or something went wrong in my PR.

[jhheider]

I need more details on why this is a bug in your opinion.

jacob@illithid brewkit % ./libexec/resolve.ts zlib.net
zlib.net=1.3.0
jacob@illithid brewkit % ./libexec/query.ts zlib.net=1.3.0 --url
brewkit: Could not fetch https://zlib.net/zlib-1.3.0.tar.gz [404]
null

using usePantry().resolve() properly populates version.raw. Using pkg.constraint.single() doesn't, resulting in the above bug.

[mxcl]

I think this is fixed with v1. Reopen if not. I traced code and couldn't see a place where .raw would be lost.