Compile-time isn't the best time for configuration.

The first argument after the list was still being replaced.
Extended test to try the hard stuff.

with a null hdr check before use (corrupt msg base?)

This reverts commit ebe52737.

This reverts commit 4ce91537.

This reverts commit 33be8662.

gitlab-runner on my Mac mini doesn't have access to /sbbs

We don't build on SunOS anymore, and using LD_RUN_PATH sucks.
Also, it was badly out of date.

attempt #2

This reverts commit 5c6a22ed.

This doesn't work cause:

Unable to create pipeline
openbsd-amd64 js-testsuite job: undefined need: openbsd-amd64 [jsexec]

and I'm not about to copy/paste those 50 lines to create the needed jsexec artifact. <sigh>

If we want to (re)run a specific set/subset of tests with jsdoor, perhaps we can
come up with some naming scheme that allows that, but for now, it's more important
that jsexec is correctly functioning than jsdoor.

Likely fixes difficulties pasting in SyncTERM.
Probably want to define some minimum amount you need to move to
trigger dragging.

The birthdate parsing is now a lot more robust and will recognize all kinds
of variants/formats and use range checks to disambiguate provided dates and
even support partial dates (e.g. year only, year/month and month/year).

The tests should all pass regardless of what the sysop has set for their
system date format in SCFG.

I had considered using sscanf() for the parsing logic but ended up going with
strtoul() only and that seems to be working well (as reflected by the passing
test cases).

... which is the format expected/required when entering or editing birthdates
in the terminal server. So this is a pretty embarassingly bad bug introduced
just before the v3.20b release to fix a less-severe bug with ecWeb new user
registration (still) writing 2-digit years to the user.birthdate field. :-(

Fix for issue #863

I considerd making these get_birthdate field functions more tolerant of less
strictly formatted strings (e.g. single-digit month or day), but decided to
keep this change to just a fix for the issue at hand. These functions could be
more robust to handle more variance in user/script input.

This is mostly as a test for using git to generate these patches.

So they're auto-triaged in Synchronet.

Also, don't do stupid things if called more than once.

This allows a server that's being terminated to stop the port-bind delay/retry
loops and terminate more quickly in cases of port binding failures.

Also not happy with this one.

Log the error string if it's non-blank, as a warning (e.g. unrecognized
text.ini text.dat string ID).

Reduce the number of logged messages when config load failure happens.

Commonize the load_cfg() error log messages.

Displayed instead of hard-coded string in xtrn.cpp when attempting to run
16-bit DOS program on system with no DOS/DOSemu support.

Displaying these settings when they're not used can confuse sysops.