[sword-devel] Bibletime 2.0.alpha3 FTBFS with sword 1.6.0RC1 library

Jonathan Marsden jmarsden at fastmail.fm
Mon Apr 13 09:08:26 MST 2009


Troy A. Griffitts wrote:

> Eeli Kaikkonen wrote:
> 
> - findConfig(&configType, &prefixPath, &configPath, &augPaths, sysconfig);
> + findConfig(&configType, &prefixPath, &configPath, &augPaths, &sysConfig);
> 
> Note BOTH the camelCase change and the addition of the &
> 
> Hope this helps.  I should have been more vocal on this change.  I
> hadn't realized anyone was using this member.  It seemed to have been
> broken in the past and so I figured it wasn't being used.

This change to BibleTime allows the build against 1.6.0RC1 to complete.

Thanks to all for the debugging assistance while I slept, and apologies
for spamming all 3 lists with this... :)

I think this incident (a) shows the value of doing test tarballs, and
(b) highlights the need for a clearly documented library API definition,
so that all developers know when an API change is happening.

In this particular instance, I can't help wondering whether there is
some way to make the API remain backward compatible (include an alias so
that both sysconfig and sysConfig are recognized as referring to that
same public data element)?  If so, then that is probably the right way
ahead; keep the API as compatible as we can, and then look at removing
the compatibility alias a couple of library releases down the road, when
all known applications have switched to using the new (correctly cased) API.

Tonight I hope to release a test BT 2.0.alpha3 package that includes
this fix as a patch, so that folks can test the BT / SWORD 1.6.0RC1
combination and see if any new issues crop up.  Presumably there will be
a BT 2.0.alpha4 before too long that includes this change.

Jonathan




More information about the sword-devel mailing list