decoder crash if delayed replication descriptors has a mistake

[vanh souvanlasy]

When a delayed replication descriptor is not followed by a 31 class descriptor. The decoder will crash instead of failing
nicely and report the error.

[vanh souvanlasy (vanh-souvanlasy)]

These BUFR message has a mistake with delayed replication descriptors sequence.
Decoding them will crash.

[vanh souvanlasy (vanh-souvanlasy)]

Fix committed in rev. 155
Now decoder no longer crash and if leave a decoding error status in decoded message flags.

[cpb (chris-beauregard)]

r155 introduces an API signature change (bufr_expand_descriptor and bufr_expand_node_descriptor get a new flag, among others). This breaks both API and binary compatibility. I don't suppose there's a solution to this problem which wouldn't require that everyone update and recompile their code?

[cpb (chris-beauregard)]

Oh, and also, bufr_expand_desc() will core dump if errflg==NULL. Should probably allow for that to make it simpler for users to update their code by just adding a NULL argument to existing bufr_expand_descriptor() calls.

[vanh souvanlasy (vanh-souvanlasy)]

Revision 158 in trunk should take care of errflg=NULL

[cpb (chris-beauregard)]

Thanks... actually, it would core if errflg was NULL and an error was actually encountered, which in some ways was nastier...