[Kst] extragear/graphics/kst/src/libkstapp
George Staikos
staikos at kde.org
Sat Jun 3 16:42:58 CEST 2006
On Friday 02 June 2006 08:56, Barth Netterfield wrote:
> AFAIR, for a while, we misused "padding" (instead of margin). Padding is
> currently undefined/buggy for legends. If you ever have padding set to
> anything other than 0, (say through scripts) you will get a borked legend.
>
> We need to maintain compatibility with .kst files saved from official
> release, and we would also like to maintain compatibility with any DR we
> released.
>
> So we can:
> -in the legend object set padding to zero until someone makes padding do
> something non-broken (put in a FIXME but not a bug - but I can see this
> biting us someday... if ever do fix padding for legends)
> -recognize file version and fix on read (don't think this is possible)
> -fix padding on legends now so it does something non-broken when it is
> non-zero (probably the best thing to do)
> -???? (something else)
>
> But I really don't think that users should have to be fixing our mistakes.
This is a dangerous road to follow. It's why the web always seems broken,
and why web browsers are always considered buggy if they're not the most
popular IE version of the time. If we put hacks like this into 1.3, then we
introduce a bug into 1.3 to compensate for a bug in 1.2. That means we have
2 bugs, and we will likely have the same one in future versions. (If all of
a sudden we support legend padding in 1.4, then geometries will be wrong.)
If we tell users that there was a bug in 1.2 and files generated with 1.2
don't look quite right in any other version of Kst, then the bug is isolated,
well-defined, and easily fixed. We could even provide a perl script to
migrate. However, there is a better solution yet, I think. As of 1.2 I
added versioning to Kst files. We can strip this at load time - your option
#2. I'll put this in now.
--
George Staikos
KDE Developer http://www.kde.org/
Staikos Computing Services Inc. http://www.staikos.net/
More information about the Kst
mailing list