What's the matter with "scm_string_p()"?

[Bruce Korb]

Hi Guys,

In trying to "upgrade" things and expunge the "gh_" interface, I
tried to use "scm_string_p()" as a replacement for "gh_string_p()".
That lead to an extraordinarily obtuse seg fault that was only
solved by using gh_string_p(), despite its being deprecated.
(Both would appropriately return #t or #f, but it seems that
scm_string_p() modified some underlying thing that triggered the
seg fault.  The actual SCM value was 11124/0x2b74.)

Anybody with any guesses as to why?  In the end, I am finding I
need to AG_SCM_WHATEVER() macro-wrap my usage of the Guile interface
so that I can upgrade for each version.  1.6.x mostly avoids the
gh_* stuff -- except for gh_string_p.

Just in case you-all wanted to know.

Regards, Bruce