Escaping backtraces makes backtrace links unclickable in TextMate

[glennfu]

This change here: b295e16

on html_formatter.rb line 103 makes it so that the backtraces are always escaped. Using the rspec textmate bundle, I'm used to running my specs in html output and then the backtrace links are clickable. Right now I have to monkeypatch this so that it doesn't escape the backtraces so I can get my clickable links back. What's the real reason for this and can it be undone?

[JonRowe]

Do you mean link's in backtraces? Or something within our formatter? The commit was to made to make the output from the formatter valid HTML.

[glennfu]

Without my monkeypatch, you can't click any of these:

With my monkeypatch, you can click these links:

[JonRowe]

I'll look into it, I won't be reverting the commit but there's obviously something going awry. I didn't notice this as an issue on the plain html formatter, is it possible this is an issue with the textmate bundle?

[glennfu]

It could be! I'll look into that myself later today.

[glennfu]

I think I'm about out of ideas of how it could be fixed through the bundle. It calls RSpec::Mate::Runner.new.run_focussed STDOUT, which then calls ::RSpec::Core::Runner.run(argv, stderr, stdout). I can't find exactly where/how it decides STDOUT then gets dumped into an html window. I hope you have better luck looking!

[myronmarston]

Ultimately this is a bug in the textmate formatter:

https://github.com/rspec/rspec-tmbundle/blob/master/Support/lib/rspec/mate/text_mate_formatter.rb

The HTML formatter isn't meant to be optimized for textmate; it's meant to be optimized to render valid HTML and it wasn't, so @JonRowe fixed it in the commit you referenced.

It looks like format_backtrace_line_for_textmate may need to be updated:

https://github.com/rspec/rspec-tmbundle/blob/024bfd2f2e731348e42126f5af37ce16896d5f2e/Support/lib/rspec/mate/text_mate_formatter.rb#L17-L22

Please open an issue there.

[glennfu]

I will do that. Thanks for helping me track that down further.

[noniq]

This was indeed a regression in 2.14, which went unnoticed because of a spec that failed to fail 😏

Pull request (against 2-14-maintenance): #1356