GitLab

Fixes to HAAN

See merge request !45

 - insecure.html as a default non-HTTPS page
 - index.html as the normal interface
 - includes an <iframe/> sourced at "/haan-generate.cgi"

 - Can also be run as a plain CGI program
 - Supports running HAAN as a backend to a web server
 - See doc/FASTCGI.MD about expectations for envvars
 - Output is a username/password combination for the SERVER_NAME
 - This requires that SERVER_NAME is setup as a HAAN virtual host

 - Did not coexist with KIP (especially, causing auxprop failures)
 - Stopped seeing KIP_REALM as an alias for HAAN_REALM; silly on a server
 - When authorization information is sought, we neutrally return SASL_NOMECH
 - Check the grammar for a user against proper size and BASE32 only
 - When a user is not found, return SASL_NOUSER
    - This is accumulated with other auxprop plugin results
    - It is only the last return when no other result is given
    - Other auxprops may be more negative or more positive, and decide

Fixes: Bugs, improvements

  * Upgraded Pypeline: Colour, Regex fork, Exit detect
  * use dynamic ports for fD client and server
  * haan uses it's own diasasl-extension-server-haan.conf
  * create sasldb2 using cmake
  * Solved a netly interaction of buffered/unbuffered
  * Set SCTP and ListenTo in freeDiameter
  * Corrected TCP:DIAMETER to SCTP:DIAMETER
  * Update of Pypeline, with SCTP improvement
  * Rename GS2-SXOVER-PLUS --> SXOVER-PLUS
  * Simplified freeDiameter test scripting

See merge request !44

 - In preparation of fixing master
 - Will build and test here first

 - Removed wrap_diameterd and used PYPE:FORK with a regular expression
 - Discovered that \s combines uglyly with CMake / CTest phases; double escapes?!?
 - Collapsed a lot of standard configuration into test macro set()tings

 - We want to derive an end-to-end key
 - This is especially interesting for 3-way communication
 - This is explicitly banned in GS2

 - Funny how habits grow, but this was never meant as a TCP port
 - Diameter is an SCTP protocol, actually one of its virtues
 - SCTP has no head-of-line blocking problems

 - SCTP is a required mode and TCP has no purpose at all
 - SCTP is _even_ available in Docker, so it must be omnipresent in servers
 - ListenTo is ignored by freeDiameter when set to localhost (::1 or the old one)
 - ListenTo is however instructive to configurations looking at this one

 - libev listens to file descriptors
 - sasl-common.c reads from buffered streams
 - diasasl-proxy-libev.c toggles between the modes
 - initial "mech>" is followed by "c2s>" with any timing
 - if the buffer loads not just the first but also the second...
 - ...libev does not receive an event for the second line!

 - Standard for CMakeLists.txt, supposedly
 - Enforced by .editorconfig

- Documentation

All problems seem resolved; now got docs and python packages

See merge request !43

 - That would only be for here & now -- which is terrible in CI
 - Rather do the general thing -- just like downloading all dependencies
 - Pack a tarball from "mkhere" and install it in /
 - Because the binhave would set it up in /tree/quickder-xxx
 - That is even true when "mkroot" is installed into /

Fixes

See merge request !42

 - <arpa2/quick-sasl.h> is now part of Quick-SASL
 - <arpa2/diasasl.h>    is now part of Quick-DiaSASL
 - <com_err/assert.h>   is now part of ARPA2 Common as <arpa2/except.h>
 - <snoprintf.h>        is now part of ARPA2 Common

This had ramifications throughout the code
 - Quick-SASL is based on membuf
 - Xover-SASL is based on dercursor
 - The two unite in things like sasl-common.c
 - A lot of patching... just to cast one type into its cousin

Add documentation through Doxygen

See merge request !41

 - Build in the GitLab-CI pipeline

 - Reuse failed on CHANGELOG... and this being master, it fails hard
 - Python was not tested; it cannot find Quick SASL anymore
 - Python needs SWIG, which the pipeline did not supply (so far trying without dependencies)
 - There are a few harmless warnings about never-occurring size wrap-arounds

 - Knowingly ignoring an irregularity in Python building
 - For details, see pipeline https://gitlab.com/arpa2/kip/-/jobs/957955273

Long-term development branch "initial" merges into "master"

See merge request !40

 - Most of it has moved over to Quick-DER
 - KIP Documents are mostly historic
 - Broke somehow, and not important anymore

Expell quicksasl diasasl

See merge request !39

Expell most ASN.1 components to Quick-DER
Be left with KIP, HAAN and support for freeDiameter,
the latter of which is scheduled to be expelled up
to a higher layer, along with the SXOVER mechanism.

 - Quite silly... message got lost in space
 - Or actually in a missing space

Finally removing DER-MEM

See merge request !38

 - The package lives on as Quick-DER / Quick-DERMEM now
 - A simpler non-DER version Quick-MEM is separately built
 - It was not actually removed before, as intended

 - Enforced on Merge Requests so new code specifies their author's rights
 - Enforced on the default branch, "master"
 - Conform https://gitlab.com/-/snippets/2058070#policy-choice-reuse-lint

no logging of binary text data

See merge request !36

Simple CI pipeline

See merge request !35