These should only be used in the sbbs_t methods called by node_thread or
event_thread. Any other thread-shared sbbs_t instances should not use these
methods since they share a single error description string buffer.

As reported by Deon on DOVE-Net: type 2.2 packet generation was not (since
SBBSecho v3.0) setting the origin and destination point values correctly:
was leaving the packet type 2.0 date values in these fields.

Also, zero-out the reserved bytes in the type 2.2 packet header (don't leave
the type 2.0 date values in there).

Thanks for the report Deon!

If the oldlen param value is less than one, set to a sane non-zero value (79).
Fix issue #576

this is recommended to reliably display the call stack when using the
print_stacktrace option.

left shift of 255 by 24 places cannot be represented in type 'int'

<phigz> what would you convert ansi to so it displays right when written out from a .js file?

Now, you can use 'ans2asc -esc' for this purpose.

e.g. running '../cleanall.sh' from the src/sbbs3 dir

As reported by GCC 10.2 with FORTIFY=1 and SANITIZE=1

Not sure how I missed this before <shrug>

This should fix the SBBS-MSVC build in CI.

This utility doesn't expand Ctrl-A codes, so just strip them.

This reverts commit a8e1ff3e.

It turns out neither platform this targeted actually set that
variable correctly.

It appears that XDG_ENVIRONMENT_TYPE will always equal "wayland"
when XWayland is in use, so when we see that, switch to CLIPBOARD
instead of PRIMARY.  This should fix WSL and ChromeOS, and be
irritating on GNOME, so it's a win all around.

The directories of the *first* library was always the target of the (new)
"Clone Settings" option. We're using a dir_t for each lib's dir_defaults, but
that doesn't mean the dir_defaults.lib is initialized to a correct libnum
(it's not), so don't use it.

Still be sure to check the return value of write() though.

The return value of x11_event() isn't actually checked anywhere currently, but
still Deuce might want to adjust the new return value here (__LINE__ or -1?).

Some indicental trailing whitespace was auto-cleaned up part of this commit.
Unexpected as I thought this code had been uncrustified already.

Should use safe_strerror() here instead... later.

Fixes recent regression with scfg

nopen.c:37:15: runtime error: left shift of 1 by 31 places cannot be
 represented in type 'int'
nopen.c:39:18: runtime error: left shift of 1 by 31 places cannot be
 represented in type 'int'
userdat.c:251:111: runtime error: left shift of 1 by 31 places cannot be
 represented in type 'int'

There are lot of 1<<31 macro definitions in this repo, so I'm expecting a
lot more of this type of fix to be required to make UBSan happy.

e.g. runtime error: store to misaligned address 0x7ffec33195e3 for type
'short int', which requires 2 byte alignment

These aren't a problem for x86, but could be a problem for some other CPU
architectures that don't support unaligned access. I didn't use memcpy() to
resolve this (though that would've worked) to make this change endian agnostic.

... declared with attribute ‘warn_unused_result’

the hash lib dependency was missing

scfgsys.c: In function ‘edit_sys_delmsg_policy’:
scfgsys.c:498:3: error: case label does not reduce to an integer constant
  498 |   case SM_USRVDELM | SM_SYSVDELM:
scfgsys.c:504:3: error: case label does not reduce to an integer constant
  504 |   case SM_SYSVDELM:

No obvious bug fixes here, just future proofing.

New issue as result signed int total_dirs (increased from uint16_t).

Because bitmap_get_scaled_win_size() was being clamped to maxsize,
no matter how much mult was incremented, w and h would never be
larger than maxwidth/maxheight and only under exceptional
circumstances would the be equal, which would result in an infinite
loop.

While we're here, set the integer scaling value so we're not chasing
ulps all over the place with floating point math/scaling when we
don't need to.

None of these were even theoretically possible on any platform, but
I took the opportunity to add bounds checking to bitmap_puttext(),
so that's good I guess.

Add some malloc() return checks, comment fall-throughs, and resolve
a sleep while holding lock.

None of these should actually matter, so hopefully I didn't add a
new bug.

This allows the ANSI yellow colour to be yellow instead of the
brown used in IBM CGA monitors.