[shiftmedia-gnutls] new port (Windows fork of GnuTLS)

[wrobelda]

Describe the pull request
Uses SMP fork of GnuTLS to add a new fork, based on the suggestion from @BillyONeal (#18029 (comment))

• What does your PR fix?

  Adds support for Windows build for gnutls. Depends on [libtasn1] Add Windows support, update to 4.17 #16953

• Which triplets are supported/not supported? Have you updated the CI baseline?

  All Windows triplets

• Does your PR follow the maintainer guide?

  Yes

• If you have added/updated a port: Have you run ./vcpkg x-add-version --all and committed the result?

  Yes

[dg0yt]

I would prefer when one port would use only one particular set of sources and patches, so that I can build for different target triplets without worrying about the state of the port's sources in the buildtrees folder.
(I know it is only a draft at this point.)

[wrobelda]

I'm sorry, I am not sure what you expect here. All of the portfiles which use ShiftMediaProject forks to provide Windows support behave like this.

[dg0yt]

I know that a lot of ports do different patching for different triplet, but this is the first time I note different sources.

• What is going to happen when I want to cross-build for Android from the same vcpkg root?
• If I need to collect the sources, am I expected to collect it separately for each triplet (read: vcpkg --download-only)?

[wrobelda]

I know that a lot of ports do different patching for different triplet, but this is the first time I note different sources.

See e.g. gmp, nettle for comparison.

• What is going to happen when I want to cross-build for Android from the same vcpkg root?

I am not sure I follow, again. Sources are downloaded into separate folders with unique names, usually consisting of a version/revision number and a hash sum. Since sources for Windows and *nix are different in those portfiles, they get downloaded into separate folders and patches are applied respectively. This behavior is no different from building a different (older) port version on demand.

[dg0yt]

Sources are downloaded into separate folders with unique names, usually consisting of a version/revision number and a hash sum.

I see, and you refering to unpacking indeed. So I only need to worry about some patches which are not applied uniformly.

Anyway, I find it confusing to have different sources. I admit I'm biased, with experience in building RPMS and DEBs.

[wrobelda]

So I only need to worry about some patches which are not applied uniformly.

I don't think there's anything to worry at all. Those Windows-specific patches will only be applied on top of ShiftMediaProject source code, which, in turn, is only used for Windows build.

Anyway, I find it confusing to have different sources. I admit I'm biased, with experience in building RPMS and DEBs.

These are definitely exceptions. ShiftMediaProject has whole bunch of MSVC-compatible forks, which are very convenient to use. I even discussed a possibility of pushing those changes upstream, but as you can imagine, it's a voluntary work on all fronts:

ShiftMediaProject/gnutls#13 (comment)

[NancyLi1013]

-- Downloading https://www.gnupg.org/ftp/gcrypt/gnutls/v3.{PACKAGE_VERSION_MINOR}/gnutls-3.6.15.tar.xz -> gnutls-3.6.15.tar.xz...
-- Downloading https://www.gnupg.org/ftp/gcrypt/gnutls/v3.{PACKAGE_VERSION_MINOR}/gnutls-3.6.15.tar.xz... Failed. Status: 22;"HTTP response code said error"
CMake Error at scripts/cmake/vcpkg_download_distfile.cmake:184 (message):
      
      Failed to download file.
      If you use a proxy, please check your proxy setting. Possible causes are:

{PACKAGE_VERSION_MINOR} should be ${{PACKAGE_VERSION_MINOR}}.

I have a question here, why do you use different repo in different platforms? Also I noticed there was a new version 3.16.6. Why do you not use the latest?

[NancyLi1013]

Please ping me if this is ready for review. @wrobelda

[NancyLi1013]

Hi @wrobelda Is work still being done for this PR? Seems there is no any progress for a long time.

[wrobelda]

Hi @wrobelda Is work still being done for this PR? Seems there is no any progress for a long time.

I am going to return to it as soon as all of my kf5* ports are merged.

[wrobelda]

@NancyLi1013 this is now ready for a review

[BillyONeal]

This approval is contingent on:

• Another vcpkg maintainer signing off on the multiple-upstreams thing
• The 'supports' expression being fixed.
• The additional quotes in file(RENAME added.

All other review comments are nitpicks.

Thanks for your contribution!

[BillyONeal]

I'm very sad to see another thing depending on yasm given how full of flaky memory corruption bugs it has turned out to be :(. No change requested.

[wrobelda]

I agree. Eventually I hope we can get a vanilla version of this to compile with clang-cl, similarly to the ongoing #26424.

[github-actions]

This is a new experimental fast check for PR issues. Please let us know if this bot is helpful!

PRs must add only one version and must not modify any published versions

When making any changes to a library, the version or port-version in vcpkg.json or CONTROL must be modified.

error: checked-in files for shiftmedia-libgnutls have changed but the version was not updated
version: 3.7.6
old SHA: 1de8a6097b77f987022f00eeeebacb163a4820ee
new SHA: 35797029a653421f3d04186ef00bc126b4d3ada1
Did you remember to update the version or port version?
Use --overwrite-version to bypass this check
***No files were updated***

[github-actions]

This is a new experimental fast check for PR issues. Please let us know if this bot is helpful!

PRs must add only one version and must not modify any published versions

When making any changes to a library, the version or port-version in vcpkg.json or CONTROL must be modified.

error: checked-in files for shiftmedia-libgnutls have changed but the version was not updated
version: 3.7.6
old SHA: 1de8a6097b77f987022f00eeeebacb163a4820ee
new SHA: 2151242268c29f3642323e0a97034a4a9c52fa09
Did you remember to update the version or port version?
Use --overwrite-version to bypass this check
***No files were updated***

[wrobelda]

This approval is contingent on:

• Another vcpkg maintainer signing off on the multiple-upstreams thing

To be fair, this is only required because vcpkg_from_github() is not currently able to download from git modules.

[BillyONeal]

Thanks!