Chokes on URL with + sign

[hasufell]

In GitLab by @ulysses4ever on Aug 5, 2022, 06:34

Doing

ghcup install ghc -u "https://gitlab.haskell.org/api/v4/projects/1/jobs/1129565/artifacts/ghc-x86_64-linux-alpine3_12-validate+fully_static.tar.xz" mr7847

I get an error:

[ Error ] Download failed: Process "curl" with arguments ["-fL", "-o",
[ ...   ]                                                 "/home/artem/.ghcup/tmp/ghcup-5295ff1124c7a668/ghc-x86_64-linux-alpine3_12-validate fully_static.tar.x.tmp",
[ ...   ]                                                 "https://gitlab.haskell.org/api/v4/projects/1/jobs/1129565/artifacts/ghc-x86_64-linux-alpine3_12-validate%20fully_static.tar.xz"] failed with exit code 22.

Notice how URL got changed: instead of + I now have a 'Reference to deleted milestone 20'.

[hasufell]

In GitLab by @maerwald on Aug 5, 2022, 09:56

Did you try with url encoding [0.1.10](https://github.com/haskell/ghcup-hs/milestone/2)B?

[hasufell]

In GitLab by @maerwald on Aug 5, 2022, 10:01

This works for me. So it's not really a bug. We just expect properly url encoded urls.

Accepting these types of URLs is an enhancement.

[hasufell]

In GitLab by @ulysses4ever on Aug 5, 2022, 10:09

Good point, thank you.

I think silently replacing + with 'Reference to deleted milestone 20' is a bug though. Correct behavior would be to abort and signal about unexpected symbol in the URL.

[hasufell]

In GitLab by @maerwald on Aug 5, 2022, 10:49

I'm on company network, so can't do much investigation now.

Check:

• https://gitlab.haskell.org/haskell/ghcup-hs/-/blob/master/app/ghcup/GHCup/OptParse/Common.hs#L221
• https://gitlab.haskell.org/haskell/ghcup-hs/-/blob/master/app/ghcup/GHCup/OptParse/Common.hs#L221

So when we read from optparse, we use parseURI strictURIParserOptions . UTF8.fromString.

When we pass to curl, we do decUTF8Safe . serializeURIRef'

uri-bytestring docs:

• https://hackage.haskell.org/package/uri-bytestring-0.3.3.1/docs/URI-ByteString.html#v:strictURIParserOptions
• https://hackage.haskell.org/package/uri-bytestring-0.3.3.1/docs/URI-ByteString.html#v:serializeURIRef-39-

[hasufell]

In GitLab by @maerwald on Aug 5, 2022, 11:00

Aha: https://github.com/Soostone/uri-bytestring/blob/1c0134cb4a15a9740d5a54807f331f511a52d1d8/src/URI/ByteString/Internal.hs#L919

[hasufell]

In GitLab by @maerwald on Aug 5, 2022, 11:03

This seems to be correct behavior: https://www.w3schools.com/tags/ref_urlencode.asp

URLs cannot contain spaces. URL encoding normally replaces a space with a plus (+) sign or with 'Reference to deleted milestone 20'.

[hasufell]

In GitLab by @maerwald on Aug 5, 2022, 11:15

Soostone/uri-bytestring#55

[hasufell]

In GitLab by @ulysses4ever on Aug 5, 2022, 11:24

Very interesting, thank you for the investigation.

I guess, feel free to close as an upstream bug…

[hasufell]

I actually changed my mind and I believe my PR fixes this: Soostone/uri-bytestring#64

RFC 3986 does not specify that + should be replaced with anything. Some HTLM encoding standards seem to specify that, but just for the query part, not the path part.