transcript runner gives confusing error when current path is missing from transcript
```ucm
> pull https://github.com/unisonweb/base:.releases._M1m .base
```
Transcript will be run on a new, empty codebase.
Initializing a new codebase in: /Users/arya/transcript-2ac5cef2900bfcb8
❓
A parsing error occurred while reading a file:
TrivialError (SourcePos {sourceName = "1682.md", sourceLine = Pos 2, sourceColumn = Pos 4} :| []) Nothing (fromList [Label ('s' :| "paces")])
It took me a minute to realize that > should have been .>.
We could improve the error message, or we could make the leading . optional when parsing a ucm block.
The code has changed a bit, but the failure is still similar. I updated the transcript to a working command (and one that won’t pull base)
```ucm
> help pull
```
and the current failure looks like
./unison-src/transcripts/fix1683.md:1:6:
|
1 | > help pull
| ^
expecting ":error", ":hide", ":hide:all", comment (delimited with “--”), end of input, or spaces
which is at least better than Show TrivialError, but still isn’t pointing at the actual issue.
This also may become valid or be handled differently once some decisions around #5173 are made.
And finally, this makes me realize that we should be correcting the source position – Megaparsec is only seeing one block at a time, so it’s assuming the start of the block is 1,1. But CMark can store its positions, and then we can tell Megaparsec what the actual starting position is in the context of the file. So I fixed that in #5387.
And with that PR, the error looks a bit better still:
Error parsing unison-src/transcripts/fix1683.md: unison-src/transcripts/fix1683.md:2:3:
|
2 | > help pull
| ^
expecting comment (delimited with “--”), end of input, or spaces
(correct line number & column, and none of the info string tags show up in the expected token list).
The transcript here is no longer invalid (as of #5826), so no longer gives an error.