In 2.0 I'd like to see include/apu_error.h simply a stub to #include <apr_errno.h> and track it all in one place. Will try to hold onto that though for my next round tuit.
It makes the back porting of new apu_errno.h constants a bit trickier but hardly impossible. ITMT, your proposal looks fine Dirk. On Thu, Jan 19, 2017 at 7:25 AM, Dirk-Willem van Gulik <di...@webweaving.org> wrote: > It seems that various error codes of what used to be apr-util do not resolve > in strings. > > Is there a larger master plan for this (where modules such as apr_crypto_FOO > can ‘register’ strings) — or would something as brutal as below > be the path for now ? > > or am I missing something. > > Dw. > > Index: misc/unix/errorcodes.c > =================================================================== > --- misc/unix/errorcodes.c (revision 1779018) > +++ misc/unix/errorcodes.c (working copy) > @@ -19,6 +19,8 @@ > #include "apr_lib.h" > #include "apr_dso.h" > > +#include "apu_errno.h" > + > #if APR_HAVE_NETDB_H > #include <netdb.h> > #endif > @@ -139,6 +141,14 @@ > return "The process is not recognized."; > case APR_EGENERAL: > return "Internal error (specific information not available)"; > +/* APR Util error codes */ > + case APR_ECRYPT: > + return "Internal error in the crypto subsystem (specific information > not available)"; > + case APR_ENOENGINE: > + return "No such engine known to the crypto subsystem"; > + case APR_EINITENGINE: > + return "Failed to init (or dlload) engine for crypto subsystem"; > + > default: > return "Error string not specified yet"; > } > >
