🐛 handle_hunk_header issue in v0.11.0

[mrjoel]

Immediately after upgrading to v0.11.0 this morning I get the following error output, so I'm reporting it. This is in my personal dotfiles repo with several working copy changes. If needed, I'll try to reduce down to a minimum reproducer for raw input.

Unexpected state in handle_hunk_header: HunkHeader(Unified, ParsedHunkHeader { code_fragment: "", line_numbers_and_hunk_lengths: [(59, 6), (59, 10)] }, "@@ -59,6 +59,10 @@", "\u{1b}[35m@@ -59,6 +59,10 @@\u{1b}[m") This should not be possible. Please report the bug at https://github.com/dandavison/delta/issues.

[mrjoel]

I encounter this warning when running a git alias I have which includes --color-words in the commandline.

I did find it notable that I get different behavior when running the command normally using pager.diff = delta where I get the warning, but not when I explicitly pipe the output into delta. In summary, for the exact working copy state:

• git diff - no issue
• git diff --color-words - shows issue
• git diff --color-words | delta - no warning reported, but display is also not as expected.

In looking a bit further, perhaps there are two separate issues at play here, the --color-words handling doesn't appear to be passing through the raw as expected from v0.10.3.

[mrjoel]

I confirmed that the --color-words handling in 0.10.3 does work as expected, so for now I've reverted back to that version.

[mrjoel]

With a bit of a closer eye and comparison, the handling of --color-words is not quite what I had expected in v0.10.3 either. I suppose it makes sense given that delta is only passing along the raw values, but for a file with multiple hunks, I'm not getting either line numbers from delta or the hunk headers from the raw git. I'd expect one or the other. Perhaps this, combined with the hunk header handling may a the toxic combination?

Note that I have my delta style using hunk-header-style = omit and line-numbers = true.

[dandavison]

Thanks @mrjoel. I agree there are two bugs here. I think I'm going to release a fix for the crash first, since that is unacceptable. But I agree that the "raw" handling of color-words isn't right either and I'll fix that subsequently.

[dandavison]

The fix for the crash is released as 0.11.1.

[mrjoel]

Thanks! Should I writeup a separate issue for the second color-words handling or would you like to? I also have a second apparent color-words handling issue I'm running into that I'll writeup.

[dandavison]

Should I writeup a separate issue for the second color-words handling

I've opened #829