Re: Field bugs was Re: commit: undebugged Lists and Fields.


Subject: Re: Field bugs was Re: commit: undebugged Lists and Fields.
From: Sam TH (sam@uchicago.edu)
Date: Wed Nov 08 2000 - 20:20:22 CST


On Thu, Nov 09, 2000 at 01:01:47PM +1100, Martin Sevior wrote:
>
>
> On Wed, 8 Nov 2000, Sam TH wrote:
>
> > On Thu, Nov 09, 2000 at 11:19:18AM +1100, Martin Sevior wrote:
> > >
> > > Undebugged lists and fields for the next release. Piece Table
> > > fields are only in the debug build.
> >
> > A couple current bugs in fields - easily visible with the test inserts:
> >
> > 1) Insert Keith's test. Note that the cursor is on the next line.
> > Type some text. It appears on the previous line, inside the
> > field. In order to get to the actual next line, press enter
> > once. The cursor doesn't move, but the actual position does.
> >
> > 2) Insert Keith's test. insert martin's test. Not only is martin's
> > test inserted in the previous line, as the previous item
> > describes, but it causes a duplicate copy of keith's test
> > to be created. This copy isn't saved, but it is displayed.
> >
> > 3) Insert Keith's test. Hit undo. Assert, then segfault.
> >
> > 4) Insert Martin's test. Hit undo. Different assert, then segfault.
> >
> > 5) Insert basically any field. Same assert as with Keith's test,
> > then segfault. *THIS IS A SHOWSTOPPER*
> >
> > I'm sure their are lots more, but this should get people started.
>
> These are not in the undebugged builds. They're in to demonstrate the
> feasibility of the piecetable fields. They only show up in
> debug builds. The non-debug builds have just our regular fields (like
> time,page number etc)
>
> If there are crash bugs associated these then they are certainly
> showstoppers.

The date&time field segfaults if you undo the insert. This is a
showstopper.

>
> To help people understand the issues, the bugs Sam reports occur
> when inserting PieceTable fields into blank documents. There are
> fewer issues when they're inserted into documents already containing text.
>
> I think the cursor on next line bug can be fixed by inserting an invisible
> run at the end of every paragraph. We can also use this run to finish the
> show paragraphs feature.

One more issue - I'm not sure I like the solution you have to putting
breaks in <c> sections. Personally, I think it would be better to
have <field ...><c>...</c><br/><c>...</c></field> than what we currently
have. However, if there are good reasons for this change, I am open
to them.

           
        sam th
        sam@uchicago.edu
        http://www.abisource.com/~sam/
        GnuPG Key:
        http://pgp5.ai.mit.edu:11371/pks/lookup?op=get&search=0xCABD33FC




This archive was generated by hypermail 2b25 : Wed Nov 08 2000 - 20:20:25 CST