Added extract_regex_match function and updated NuGetProvider

[kazssym]

A new function extract_regex_match has been added to the vcpkg namespace in tools.cpp. This function performs a regex search on an output string and returns the first match or a formatted error message.

The NuGetProvider struct has been updated to replace the extract_prefixed_nonwhitespace function call with the new extract_regex_match function. This change enhances the flexibility and robustness of version extraction by using a regex match instead of a fixed prefix.

Fixes microsoft/vcpkg#38940

[kazssym]

I tested locally with the Japanese localization. Please confirm it is OK.

[BillyONeal]

The whole reason extract_prefixed_nonwhitespace exists is that we are trying to break the dependency on std::regex because all major implementations of that are awful. Only the 'fake XML parsing' thing still uses it, and that entire component's days are numbered as there is no longer any reason to have a separate vcpkgTools.xml now that the bash bootstrap script no longer looks at it.

Can you write code that implements this instead and a comment with the example non-English output with which it is expected to work?

Alternately, should we be telling NuGet we want locale-independent output for this test somehow?

[kazssym]

The whole reason extract_prefixed_nonwhitespace exists is that we are trying to break the dependency on std::regex because all major implementations of that are awful. Only the 'fake XML parsing' thing still uses it, and that entire component's days are numbered as there is no longer any reason to have a separate vcpkgTools.xml now that the bash bootstrap script no longer looks at it.

Can you write code that implements this instead and a comment with the example non-English output with which it is expected to work?

Alternately, should we be telling NuGet we want locale-independent output for this test somehow?

Thank you. The issue this pull request tries to resolve is nuget.exe's output is now locale-dependent and no longer has a fixed prefix. A sample is NuGet バージョン: 6.10.0.107 with the Japanese locale.

Possible alternatives would be either

1. Execute nuget.exe always with the English locale.
2. Use pattern matching without regex.
3. Use a consistent regex engine.

Which one is better?

[kazssym]

I found the -ForceEnglishOutput option can be used to get fixed locale-independent outputs, but it could not be used without any subcommand as in nuget.exe -ForceEnglilshOutput.

[kazssym]

Setting the NUGET_CLI_LANGUAGE environment variable seems OK. Can we set this variable when executing nuget.exe?

[myd7349]

It works for me:

nuget.exe help -ForceEnglishOutput

[kazssym]

It works for me:

nuget.exe help -ForceEnglishOutput

Hmm, you used the help command while I didn't.

[kazssym]

Rewrote this PR in #1451.

[BillyONeal]

Thanks!