[BUG]DVS breaks FileContentsBrowser diffing
Colin Putney
cputney at wiresong.ca
Fri Jun 27 20:58:40 UTC 2003
Ned suggested collapsing these character sequences into a single CR:
CR
CRLF
CRCRLF
CRLFLF
On Friday, June 27, 2003, at 01:16 PM, Julian Fitzell wrote:
> Hrm... what makes the last few a single CR by definition instead of
> two CR's ? But I agree that getting something consistent out of it
> would be good in one way or another.
Ned responded:
> Because the last two are not valid, and are the result of breaking the
> files with older bugs. They started out as single line endings and
> were broken. They should be read in as single line endings.
> There have already been such broken files checked into the aubergines
> CVS.
I think the correct approach is the fix the files and fix the tool so
that it doesn't break any more files. If we try to make the tool
automatically detect and fix broken files, we'll run into two problems.
1. We might mask a bug in the output code.
2. We might introduce new forms of corruption. Consider a method with a
string literal containing CRLF. Ok, that's a strange thing to do, but
it should be possible to file that method out and read it in again.
I do think we've finally got Monticello where it needs to be. If any
really exotic use cases come up we can add them to the test suite and
go from there.
Colin
More information about the Squeak-dev
mailing list
|