Log errors if/when comWriteBuf() fails.

This is to hopefully help to root-cause or even possibly resolve the issue
reported in #synchronet by PyRoDrAkE:
"It all works well until the Dialup connection runs slower than 4800 baud
using SexPots. I connect from my XT at either 1200 baud or 2400 baud and it
would just stop drawing midway."

There shouldn't be any write timeout in comWriteBuf(), so I don't know why
this would help, but it's worth a try. Logging any comWriteBuf() errors is
also a good idea in general.

and if the message is UTF-8, encode the origin line as UTF-8 too.
If the message is straight ASCII (e.g. the sub-board is configured as ASCII
only), convert the configured origin line to ASCII, before appending.

At this time, only ASCII and CP437 encodings of originlines are supported
in Synchronet (imported and exported origin lines can contain UTF-8).

This should address the issue raise by Sergey Dorofeev (2:5020/12000) in the
FidoNet UTF-8 echo whereby messages exported from Synchronet systems
(presumably using SBBSecho) could contain a mixture of CP437 and UTF-8
encodings (!) when the sysop had CP437 chars in their configured origin line.

I sent an update ("add" email), but it was rejected with
"No LSM entries were found" - whatever that means.

... so yeah, most *nix sysops will get a warning about running something newer
than the latest release.

... for a version number comparison (e.g. by chksetup.js)

This file hasn't been updated in 20 years and nobody noticed. :-)

I at least made a new submission (sent email) one last time before we
(might) just remove this file since the the LSM database doesn't appear
to be of any real use/value these days.

This fixes the bug whereby if the user ever hit Ctrl-C or hit 'N' at
a pause prompt, they could never see another menu / display file.

When displaying this node vstatus, the user's alias shouldn't be appended to
the node status, else you get something like this:

16  Logging out Digital ManDigital Man

It was possibly before for fgets() failures to be pretty hard to diagnose.

But set the error string to hopefully be printed as a warning

A sysop could hit 'q' to quit (sbbs, console mode) while the binding is taking
place.

Fixes assertion on failed connects.

Try for 16 seconds (14 loops) to load the SSL certificate.

On the first time through the loop, create self-signed certificate
if configured to do so.

This also splits the generation of the self-signed certificate into
a separate function.

While we're here, split the new SSL epoch out into a separate function
as well, and explicitly call it when we create a new self-signed
cert. This at least partially fixes the epoch thing, but there's
still the possibility of creating it multiple times in the same
second... the file date isn't really enough.

For quick open/write/close log files (e.g. error.log), go back to using
nopen() to insure exclusive access while writing to prevent any accidental
corruption from concurrent writes from multiple threads or processes.

Default to false / No. Maybe this should default to true / Yes until a
certificate (e.g. from Let's Encrypt) is successfully provisioned?

Anyway, this is just a quick-hack to address the issue where a sysop's
CA-sigend certificate may be overwritten with a self-signed certificate if
for some reason Synchronet can't find/read the ssl.cert file.

See issue #881.

Missing WINAPI decoration on function loaded from DLL.
When called, this would corrupt the parameters to callers,
presumably because the register allocater didn't know the call
would tromp on something.

This reverts commit 219e6858.

This reverts commit 4f2886fc.

This should be redundant since "zip" and "unzip" should already be in the
native program list, but at some point we should be able to get rid of the
native program list or at least have it default to blank.

Cryptlib uses the NT 3.1 API for ACLs, which means hard-codeding
certain security structure sizes.  Quadruple the sizes of these
to see if this fixes the crash DigitalMan is seeing.

Unlike SMB_MSGS, the sub-board does not have to be already open.

As requested/reported by Finglonger (SILENT)

Don't add the atexit() do_cryptEnd() handler until we're not going
to call cryptEnd() ourselves.

<@Deuce> If you want the install test to start immediately, you would use needs: []

so they can run sooner in the pipeline

Possibly should be done after a failure too...

DDFileLister: Fix for editing a file's extended description when the file doesn't have one initially

See merge request !519

Eric Oulashin authored 1 month ago and Rob Swindell committed 1 month ago

DDFileLister: Fix for editing a file's extended description when the file doesn't have one initially

Also, fix some formatting.

Fix for issue #884

We must wait until the connect_callback() has been called to publish these
messages. Using mqtt.server_version as the indicator of an initial startup
connection versus a re-connection (due to broker-connection loss), i.e.
setting to NULL after initial connection.