fix external editor for cmd and bat on windows

[pm100]

fix for #1316

resubmitting after commit cleanup

the actual fix is

• if execution fails on windows
• attempt to rerun as cmd /c <cmd> <args> <file name>
• if that fails then report error

verified with both vim (which is installed a bat file) from the original report and code (which is installed as a cmd file) which was mentioned in a further comment

It is arguable that rust's program exec function should do this itself. But is also arguable that it shouldnt. I will file an issue there

Then I added a bunch of tests. One is generic, the other test this specific fix. I could maybe add more later.

In order to make the tests work I need code from the test plumbing in the asyncgit sub crate. There are 2 choices to do that, given that code flagged as test on one sub crate is not available to other crates in the project.

• duplicate the test infrastructure code
• move it to a 'pseudo' feature for test only - see make code compiled with cfg(test) visible across crates rust-lang/cargo#8379

I chose the second.

Sadly this makes clippy and cargo check see that plumbing code (that is only compiled into tests) as part of the main project. They complain big time about that code. I fixed some complaints, I told clippy to ignore some, but the main (not real issue) is that they see duplicated packages. It is almost impossible to disentangle them. And its not possible to tell clippy and check to skip that fake feature. So i added a skip duplicates check (the same as the main project) and modified deny.toml

I followed the checklist:

• [ x] I added unittests
• [ x] I ran make check without errors
• [ x] I tested the overall application
• I added an appropriate item to the changelog

[extrawurst]

so far I managed to keep duplicate versions of syn away. can we achieve this here too?

[pm100]

its only testing - the dup version is needed by test serial. You have the same testing issue when you use test serial but the validation tools dont see it. BUT due to how I pull asyncgit into testing in the main tree it seems to the various tools like this is a dup in the real code, its not. (I commented earlier about using asyncgit code in tests)

I can take out the tests, that solves all.

Or I can duplicate the test scaffolding from asyncgit into the main tree, then the tools wont see it.

Or I can upgrade most things to the same syn (v2) but I would have to pull local versions of three things into gitui tree since they are marked syn=1.x

[extrawurst]

if there is a way to migrate to syn v2 this would be my preferred solution. this should happen in a separate PR though. can you give this a try? this will set things up for the future

[pm100]

after cargo update we are left with three things using 1

pm100@pauls-MacBook-Air gitui % cargo tree -i syn@1.0.109 --all-features
syn v1.0.109
├── git-version-macro v0.3.5 (proc-macro)
│   └── git-version v0.3.5
│       └── bugreport v0.5.0
│           └── gitui v0.22.1 (/Users/pm100/oops2/gitui)
├── proc-macro-error v1.0.4
│   └── struct-patch-derive v0.2.3 (proc-macro)
│       └── struct-patch v0.2.3
│           └── gitui v0.22.1 (/Users/pm100/oops2/gitui)
└── struct-patch-derive v0.2.3 (proc-macro) (*)
pm100@pauls-MacBook-Air gitui %

proc-macro-error has 2 PRs open upgrading syn to v2.
git-version-macro has 3 month old PR too
I opened an issue for struct-patch-derive

[pm100]

I have resolved this by getting everybody onto syn v1. Once all the dependencies move to v2 then gitui can move.

I also had to fix a new clippy failure due to a new rule added in 1.71

[extrawurst]

depends on #1781

[extrawurst]

since #1781 is closed now, can we update this to current master and move it forward?