Mailing List Archive

I proposed the monster mod_cgi/cgid unification backport for 2.4.x, but
wanted to give a bit more context here rather than in STATUS. PR is
here: https://github.com/apache/httpd/pull/209

1) stderr handling is a significant regression in mod_cgid compared to
mod_cgi so it is worth fixing this IMO (PR 54221)

2) we have shipped this patch in Fedora&RHEL for a few years, so I am
quite confident that the file descriptor passing feature works well *on
Linux*. I have no idea how well it works on non-Linux platforms, hence
keeping it opt-in (./configure --enable-cgid-fdpassing).

3) there should be no behaviour changes to mod_cgi or mod_cgid if the
fdpassing feature is not enabled, except where bug fixes to either were
missing and are provided by the unification of common code.

4) because the bulk of the patch is de-duplication, this is a net
reduction in lines of code shipped despite the new fd-passing code.
diffstat:

.github/workflows/linux.yml | 8 -
changes-entries/pr54221.txt | 3
modules/generators/cgi_common.h | 639 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
modules/generators/config5.m4 | 11 +
modules/generators/mod_cgi.c | 586 ++++----------------------------------------------------------------------------------------------------
modules/generators/mod_cgid.c | 642 +++++++++++++++++++++++++++++++++++++++++-------------------------------------------------------------------------
6 files changed, 913 insertions(+), 976 deletions(-)

Since the github PR link is a series of patches, attached the complete
patch here as well in case that's preferred for review.

Regards, Joe