Crashes on long path lengths on Windows

[epage]

cargo just got a report of a crash on Windows with long paths (rust-lang/cargo#11710)

Problem

cargo clean, cargo tree, cargo build, etc. generates panic

thread 'main' panicked at 'range end index 270 out of range for slice of length 260', C:\Users\runneradmin\.cargo\registry\src\git.luolix.top-1ecc6299db9ec823\is-terminal-0.4.0\src\lib.rs:169:14 stack backtrace: 0: 0x7ff6d75f5db2 - git_midx_writer_dump 1: 0x7ff6d761b89b - git_midx_writer_dump 2: 0x7ff6d75ecb4a - git_midx_writer_dump 3: 0x7ff6d75f5afb - git_midx_writer_dump 4: 0x7ff6d75f8ef9 - git_midx_writer_dump 5: 0x7ff6d75f8b7b - git_midx_writer_dump 6: 0x7ff6d75f9791 - git_midx_writer_dump 7: 0x7ff6d75f951e - git_midx_writer_dump 8: 0x7ff6d75f6aaf - git_midx_writer_dump 9: 0x7ff6d75f91d0 - git_midx_writer_dump 10: 0x7ff6d76ae435 - git_midx_writer_dump 11: 0x7ff6d761e267 - git_midx_writer_dump 12: 0x7ff6d76ae959 - git_midx_writer_dump 13: 0x7ff6d74a4659 - git_midx_writer_dump 14: 0x7ff6d6cc7087 - git_libgit2_prerelease 15: 0x7ff6d6cc15ba - git_libgit2_prerelease 16: 0x7ff6d6cc1499 - git_libgit2_prerelease 17: 0x7ff6d6c3f8c7 - git_libgit2_prerelease 18: 0x7ff6d6cb77be - git_libgit2_prerelease 19: 0x7ff6d6c34e16 - git_libgit2_prerelease 20: 0x7ff6d6c44c2c - git_libgit2_prerelease 21: 0x7ff6d75e4f0e - git_midx_writer_dump 22: 0x7ff6d6cba51c - git_libgit2_prerelease 23: 0x7ff6d7624ee0 - git_midx_writer_dump 24: 0x7ffb5eb184d4 - BaseThreadInitThunk 25: 0x7ffb60ca1791 - RtlUserThreadStart

Steps

1. setup rust build in remote session, Jenkins, for example, using really long file names

2. use next script to be called by CI

rustup default 1.67.0
rustup update 1.67.0
cargo clean

3. panic

Possible Solution(s)

Switch to version 1.66.0

Notes

Normal Windows command line Session works fine.

Problem lies in is_terminal:0.4.0, which failed if real terminal is not available. The problem is related to the badly supported long file names: 260 is a MAX_PATH, and we have really long paths. thread 'main' panicked at 'range end index 270 out of range for slice of length 260' code```

 let mut name_info = FILE_NAME_INFO {
    FileNameLength: 0,
    FileName: [0; MAX_PATH as usize], //// <- ERROR FileNameLength is 260
};
// Safety: function has no invariants. an invalid handle id will cause
//         GetFileInformationByHandleEx to return an error
let handle = GetStdHandle(fd);
// Safety: handle is valid, and buffer length is fixed
let res = GetFileInformationByHandleEx(
    handle,
    FileNameInfo,
    &mut name_info as *mut _ as *mut c_void,
    std::mem::size_of::<FILE_NAME_INFO>() as u32,
);
if res == 0 {
    return false;
}
let s = &name_info.FileName[..name_info.FileNameLength as usize]; // <- FAULT FileNameLength is 270

### Version

```text
cargo 1.67.1 (8ecd4f20a 2023-01-10)
release: 1.67.1
commit-hash: 8ecd4f20a9efb626975ac18a016d480dc7183d9b
commit-date: 2023-01-10
host: x86_64-pc-windows-msvc
libgit2: 1.5.0 (sys:0.16.0 vendored)
libcurl: 7.86.0-DEV (sys:0.4.59+curl-7.86.0 vendored ssl:Schannel)
os: Windows 10.0.14393 (Windows Server 2016 Standard) [64-bit]

[ChrisDenton]

That seems weird. The code in question checks for failure, and not having enough space in the buffer should be a failure:

is-terminal/src/lib.rs

Lines 158 to 166 in bee73f2

	let res = GetFileInformationByHandleEx(
	handle,
	FileNameInfo,
	&mut name_info as *mut _ as *mut c_void,
	std::mem::size_of::<FILE_NAME_INFO>() as u32,
	);
	if res == 0 {
	return false;
	}

Maybe it's a result of a server that's using customized drivers (or other third party kernel tweaks) or maybe there's a bug in that version of Windows Server. Either way I guess it'll have to verify the FileNameLength too.

[sunfishcode]

I've now prepared #19 with a check that the filename length is in bounds before we use it.

[weihanglo]

Hi @sunfishcode, is there any plan to have a point release in a few days? We're going to cut a beta backport on Cargo side. No worries if there is not. Thank you for the quick fix!

[sunfishcode]

I've now released is-terminal 0.4.4 with this fix.