httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Roy T. Fielding" <field...@kiwi.ics.uci.edu>
Subject Re: Global Renaming: Let us start!
Date Mon, 06 Apr 1998 00:00:21 GMT
I was just going to post the same comments (well, some of them).

>Ok I don't understand what the Big Change is, it's not documented in
>STATUS... and there's like 50 messages on the topic that I was hoping were
>summarized somewhere. 

The big change would be

    tag PRE_AP_PREFIX_RENAME
    make rename
    compile entire server (without the back-compat)
    check symbols
    check operation of server
    compile entire server with the back-compat
    check symbols
    check operation of server
    commit
    tag POST_AP_PREFIX_RENAME

A minor problem I've had so far is that I can't compile
mod_auth_db on Solaris (before changing anything).

A major problem is that I can't compile csubst either:

   % make rename
   make[1]: Entering directory `/extra/fielding0/private/ws/ap3/src/test/rename/csubst'
   gcc -O2 -I. -c csubst.c
   csubst.c:6: err.h: No such file or directory
   make[1]: *** [csubst.o] Error 1

Ralf, do I need the idutils to compile this, or is err.h supposed to be
a system header file?


>There is src/test/rename/README but it doesn't say what solution it was
>designed to implement.  and apapi.h in that directory has some
>questionable stuff: 
>
>extern ;
>
>and
>
>/* prototype for `ap_construct_server' cannot be found */
>
>but:
>
>% rgrep construct_server
>./main/util.c:API_EXPORT(char *) construct_server(pool *p, const char *hostnam
>e,
>./test/rename/apapi.h:/* prototype for `ap_construct_server' cannot be found *
>/
>./test/rename/apapi_compat.h:#define construct_server               ap_constru
>ct_server
>./include/httpd.h:API_EXPORT(char *) construct_server(pool *p, const char *hos
>tname,
>./include/hide.h:#define construct_server               AP_construct_server
>
>Does it maybe not understand multiline prototypes?

That's what I figured.  Actually, it's a miracle that it can do as much
as it does.  I'll hand-patch those if needed.

>In apapi_compat.h there's the comment:
>
>/*
> *   Mapping of Apache 1.2 symbols to
> *   Apache 1.3 symbols which are
> *   _NO LONGER PART OF THE API_
> */
>
>Which isn't true... those are CORE_EXPORT symbols, which are still
>part of the private API... they were never part of the public API.
>The comment should probably read "mapping of symbols which are private
>to httpd and should not be considered part of the API".
>
>The following are incorrectly placed into the DO_NOT_USE section:
>
>#define bprintf                        DO_NOT_USE_bprintf
>#define bvputs                         DO_NOT_USE_bvputs
>#define day_snames                     DO_NOT_USE_day_snames
>#define fnmatch                        DO_NOT_USE_fnmatch
>#define is_fnmatch                     DO_NOT_USE_is_fnmatch
>#define month_snames                   DO_NOT_USE_month_snames
>#define null_cleanup                   DO_NOT_USE_null_cleanup
>#define os_is_path_absolute            DO_NOT_USE_os_is_path_absolute
>#define parse_uri                      DO_NOT_USE_parse_uri
>#define psprintf                       DO_NOT_USE_psprintf
>#define pstrcat                        DO_NOT_USE_pstrcat
>#define register_other_child           DO_NOT_USE_register_other_child
>#define rvputs                         DO_NOT_USE_rvputs
>#define send_header_field              DO_NOT_USE_send_header_field
>#define server_root                    DO_NOT_USE_server_root
>#define set_file_slot                  DO_NOT_USE_set_file_slot
>#define set_flag_slot                  DO_NOT_USE_set_flag_slot
>#define set_string_slot                DO_NOT_USE_set_string_slot
>#define set_string_slot_lower          DO_NOT_USE_set_string_slot_lower
>#define table_do                       DO_NOT_USE_table_do
>#define unregister_other_child         DO_NOT_USE_unregister_other_child
>
>Many of these are API_EXPORT_NONSTD and so that's why I guess they're
>messed up.

Hmmm, we should fix the build script for those.  Ralf?

>It looks like check_alarm() needs to be part of the API -- otherwise
>folks can't write win32 modules that do timed stuff (except timed stuff
>that writes through a BUFF.)

Agreed.

....Roy

Mime
View raw message