[darcs-users] Re: more feedback on binary file detection (issue pin-pointed?)

[Mark Stosberg]

I have another data point on binary file detection. This one points back
to darcs. :)

I opened a file and saw ^@ at the bottom and removed it.

I tried to record it, and it was detected as binary. I used two tools
(khexedit and the 'tr' command below) to confirm it was (now) free of
NUL and EOF.

 tr -dc '\000\012\032' <../../sql/data_model.sql | tr '\000\010\032' '@\010Z' | grep -n .

I checked 'pending' to make sure it was empty. 

I tried 'record' again and darcs detected the file as binary, despite the lack
of binary characters in the file. 

My suspicion is that it is being considered (wrongly) binary because it a version
with a trigger character exists in 'current'.

[ tries something ]. 

I think I was just able to confirm this. I went ahead and recorded the
file as binary. I immediately tried to record it again, and it detected
the file as text. 

This behavior indicates the the copy of the file in 'current' rather
than the working directory is used to check for binary-ness. 

Perhaps this update would be an easy change for an Haskell programmer to make. :)

    Mark