Diff in the web interface

[savatin]

• Gitea version (or commit ref):
  1.14.0+dev-18-gd453533be
• Git version:
  2.26.2
• Operating system:
  Ubuntu 18.04.3 LTS
• Database (use [x]):
  • PostgreSQL
  • MySQL
  • MSSQL
  • SQLite
• Can you reproduce the bug at https://try.gitea.io:
  • Yes (provide example URL)
  • No
• Log gist:

Description

The ' and " characters are displayed incorrectly when diff files are displayed in the web interface:
https://try.gitea.io/sergey.savatin/test_diff/commit/62d96b3486efecfce9736c7c3382077c10bfa6c4

Screenshots

[silverwind]

HTML source looks rather broken too with a 0x03 character in there:

<span class="mono wrap">    sh �<span class="added-code">9;useradd -u $(stat -c "%u" .gitignore) jenkins'</span>;</span>

[zeripath]

The issue is that the diff library appears to be diffing the html entities at the subentity level.

so " is being compared with ' which means you get &#3 separated from the 9; one side and 4; on the other.

[zeripath]

I think we should just be letting chroma do the highlighting by passing in a HighlightRanges() and adjusting the format that chroma.LineHighlight sets.

[zeripath]

I'm too tired to do this right now but the idea would be rewrite gitdiff.go GetComputedInlineDiffFor to generate the line patches first and then pass to another highlight.Code - which would do the addition and deletion bits with the appropriate bits highlighted by calling the HighlightRanges and adjust LineHighlight options as appropriate.

[zeripath]

(@mrsdizzie you don't have to fix - but if you understand what I'm getting at feel free to have a go.)

[mrsdizzie]

Grr yea : ( I'm away for a few days but could take a look after

[zeripath]

ugh the highlight trick suggested above doesn't work.

[zeripath]

OK I worked it - we just have to iterate through the diffs and run a fix up for the broken entities.

Now doing this I note that we end up creating these line diffs twice - which seems excessive to me - so clearly there are improvements to be made here.