Pasting code snippets (especially diffs) now converts it to markdown, which completely messes up the text.
What version of the Codex App are you using (From “About Codex” dialog)?
26.707.91948
What subscription do you have?
Pro
What platform is your computer?
Darwin 23.6.0 arm64 arm
What issue are you seeing?
When I'm code reviewing changes made by Codex, I often copy/paste parts of the diff to provide context for my requested change, question, or whatever. This used to work great. Now (after the conversion from Codex to ChatGPT) it's started automatically converting these into markdown, which is completely inappropriate and totally screws up the formatting. Please either revert this anti-feature or (more likely I presume), add a way to turn it off. Let me go back to just typing/pasting as plain text. Not markdown.
What steps can reproduce the bug?
Paste something like:
@@ -347,7 +354,13 @@ internal fun LiveGameScreen(
}
},
onLock = { locked = true },
- onGoal = { team -> onStateChange(state.recordGoalFromCurrentState(team, now)) },
+ onGoal = { team ->
+ val updatedState = state.recordGoalFromCurrentState(team, now)
+ onStateChange(updatedState)
+ if (updatedState.automaticWaterBreakReached(team)) {
+ showWaterBreakPrompt = true
+ }
+ },What is the expected behavior?
All those + and - lines should not turn into bullets. It should just paste the text as given without trying to convert to markdown.
Additional information
Codex itself thought I should be able to paste inside a [three backtick] region, but that doesn't work either. Here is what the above turns into when I do its suggestions ([three backticks]diff first, then Cmd+Shift+V to paste):
@@ -347,7 +354,13 @@ internal fun LiveGameScreen(
}
},
onLock = { locked = true },
- onGoal = { team -> onStateChange(state.recordGoalFromCurrentState(team, now)) },
- onGoal = { team ->
- val updatedState = state.recordGoalFromCurrentState(team, now)
- onStateChange(updatedState)
- if (updatedState.automaticWaterBreakReached(team)) {
- showWaterBreakPrompt = true
- }
- },All the + turned into - and the indentation was swallowed. So it seems to be converting to markdown and then back to text through this process. This is unacceptable formatting.
7 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
This is just insane. They keep forcing "features" that:
A. Do not work
B. Completely unnecessary for the 90% of users, namely, those who uses Codex for code-related tasks.
See #34071 for more examples of how that "auto markdown" messes with text we just try to write (sic!)
Such a showstopper. The agent doesn't seem to know what line you have selected, copy-pasting is the most natural and established way and now... you can't even paste the code without first converting it to plain text?
More importantly, you literally cannot reliably convert it to a form that will be consumed as plain text because, in fact, the composer treats any plain text as markdown. If your text has
-,+, or*at the start of a line, it will be treated as a markdown bullet, regardless of what it is in reality.But, indeed, converting to plain text manually first is still necessary when you paste code copied from VSCode if you use any theme that has italics (such as Dracula or Night Owl), see #33586.
It's terrible—I can't close it! I even have to take screenshots just to keep track of the information I need to enter.
The latest version at least pastes correctly within three backtick regions. But that shouldn't be required. I still consider this a bug that needs to be fixed. There should be an option to turn off automatic markdown entirely in the prompts. It's not like Codex can't understand the raw markdown. If you want to make a bulleted list, it understands raw * or - lists just fine. No need for the harness to convert things before it gets to the AI.
@rmjarvis They likely just don't care about anything much. QA for all Codex UIs is very bad since the beginning.