inaccurate version bounds in .cabal file

[hvr]

As can be seen on http://matrix.hackage.haskell.org/package/second-transfer several releases of second-transfer are suffering from bit-rotting install-plans due to omitted upper bounds or inaccurate version bounds in general.

[alcidesv]

Fixed!

[hvr]

@alcidesv I'm afraid it's not fixed yet completely:

Configuring second-transfer-0.10.0.4...
Preprocessing library second-transfer-0.10.0.4...

hs-src/SecondTransfer/TLS/CoreServer.hs:3:14:
    Unsupported extension: PartialTypeSignatures

so that file should have been listed as other-extensions: in the .cabal file so the cabal solver knows which features the Haskell compiler needs to provide, and even if that one was adressed:

[ 1 of 47] Compiling SecondTransfer.Http2.CalmState ( hs-src/SecondTransfer/Http2/CalmState.hs, /tmp/second-transfer-0.10.0.4/dist-newstyle/build/second-transfer-0.10.0.4/build/SecondTransfer/Http2/CalmState.o )

hs-src/SecondTransfer/Http2/CalmState.hs:16:28:
    Not in scope: type constructor or class ‘Word’

hs-src/SecondTransfer/Http2/CalmState.hs:16:34:
    Not in scope: type constructor or class ‘Word’

which means that the code doesn't work with base-4.7 anyway. So the lower bound on base is still inaccurate.

I'm sure there are other inaccuracies in the .cabal file which will become an issue sooner or later.

Morever, somebody will have to go over all past second-transfer releases and add missing version bounds, as it's not enough to just release a new version of second-transfer with tighter version bounds. The cabal solver will happily choose an older release of second-transfer with less tight version bounds.

[alcidesv]

Thanks @hvr . This is mostly due to my lack of experience, I will be fixing the bounds issue. I didn't know about hackage matrix and about other-extensions. How can I get the "Unsupported extension" messages in my build?