Redefine PATH_MAX to allow all Git operations to take advantage of Windows long path support.

[lucasz93]

PATH_MAX is inherited from mingw64/include/limits.h, with a value of 260.
The problem is that even after enabling Windows long path support and 'core.longpaths' there are still some Git operations that will fail, such as cloning a submodule (see setup_explicit_git_dir).

I've observed some CI runners that have temp clones in a deeply nested path which will break even some medium sized repository paths.

Redefining this macro addresses that issue.

This does seem very hacky, but it was a big issue for my Windows Bitbucket runner which clones into a ridiculously long path of: C:/Windows/System32/atlassian-bitbucket-pipelines-runner/temp/b00863ba-0d50-5cf5-91e6-452e9580164b/1709559465808/build. This caused some of my dependent submodules to fail initialisation.

Examples of the same issue:
aws/aws-iot-device-sdk-cpp-v2#157
gruntwork-io/terragrunt#2343

If there's another possible way to implement this - I'm open to ideas. Would just like to see this issue go away :)

[dscho]

As mentioned below, the commit message needs improving, because it needs to convince the reader that this is a desirable change. A couple of important things are still unclear to me:

• Did you mean to mention Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\FileSystem\LongPathsEnabled in the commit message? I think it would be highly advisable to do so.
• What is the PATH_MAX value going to be if we #undef it?
• What is the consequence of running this with a long path when core.longPaths=true and Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\FileSystem\LongPathsEnabled = 0?
• What is the consequence of running this with a long path when core.longPaths=true and Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\FileSystem\LongPathsEnabled = 1?
• What is the consequence of running this with a long path when core.longPaths=false and Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\FileSystem\LongPathsEnabled = 0?
• What is the consequence of running this with a long path when core.longPaths=false and Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\FileSystem\LongPathsEnabled = 1?

[dscho]

I can understand most of this, but since you are contributing this to the Git for Windows project, I don't think it would be appropriate to include the _CYGWIN part.

Also, the commit message is a bit too mum about rationale. Please follow the guidance in https://github.blog/2022-06-30-write-better-commits-build-better-projects/ to improve it, in particular with a strong focus on this part:

	What you’re doing	Why you’re doing it
High-level (strategic)	Intent (what does this accomplish?)	Context (why does the code do what it does now?)
Low-level (tactical)	Implementation (what did you do to accomplish your goal?)	Justification (why is this change being made?)

[rimrul]

What is the PATH_MAX value going to be if we #undef it?

4096

[dscho]

What is the PATH_MAX value going to be if we #undef it?

4096

Right. Thank you for looking it up, I vaguely remembered something like that.

However, my point was that the commit message needs improving.

But talking back and forth between reviewers won't fix that.

[dscho]

@lucasz93 feel free to respond to the reviews. We can reopen this PR then.

[adalmia10]

@dscho
I am getting the same issue that this PR tried to fix, are you aware of any alternative solution?

[dscho]

are you aware of any alternative solution?

No.