Diagnosing this issue was made a bit harder by the fact that the configure script unconditionally writes
Gz.................. no (zlib is too old)
in pike_Gz.feature. Shouldn't the
PIKE_FEATURE(Gz,[no (zlib is too old)])
code only be executed if the result of the AC_TRY_LINK above fails?
Normally, when a working libz is found, a new PIKE_FEATURE is executed
by AC_CHECK_GZ. But there is no other PIKE_FEATURE if no working libz
is found. That seems to be a separate, less important, bug.
