patch: crash on libddcutil init with Philips 328E1 about ddcutil HOT 4 CLOSED

Thank you for the patch. Do you have a ddcutilrc configuration file? If so what is its contents?

from ddcutil.

Yes, I had this ~/.config/ddcutil/ddcutilrc file when the error occurred:

[global]
options: --sleep-less --sleep-multiplier .025 --timeout-i2c-io

[ddcutil]

[libddcutil]
--noverify

When I remove this config file, ddcutil git commit f2bc353 works fine without my patch.
I don't need the config file anyway. My app calls

ddca_enable_report_ddc_errors(false);
ddca_enable_verify(false);
ddca_enable_sleep_suppression(true); // I know this is a noop nowadays

on startup, and then does this in an async thread that processes queued change requests in a loop:

__useconds_t sleep_usec = 45000; // Philips 328E1: 45000 always works, 44000 occasionally drops a packet
for (;;) {
    ...
    DDCA_Status rc = ddca_set_non_table_vcp_value(ddca_dh,code,v >> 8,v & 0xff);
    if (rc) error(...);
    usleep(sleep_usec);
}

Sorry for not having thought about the config file myself.

from ddcutil.

I have pushed a fix for the crash you reported to branch 1.4.0-dev. The underlying problem was that function apply_config_file() in file ddcutil_config_file.c was not setting the proper return values for the case where the config file is invalid. The assert() failure in function ntsa_length() is a precondition failure.

The configuration file error that triggered the problem is the [libddcutil] segment, which should read "options: --noverify",
not "--noverify".

Thank you for reporting the problem. Getting all the bugs out of error handling code is difficult, since execution paths through error handlers are infrequent.

I have also modified the libddcutil initialization code to write the error message to the system log as well as syserr, since messages to syserr can be lost depending on how the program using the shared library is invoked. As the libddcutil initialization function __ddca_init() is a constructor it doesn't set the thread error detail that can be checked by the caller as with other API functions. Perhaps there needs to be an additional API function ddca_get_initialization_msgs() to get this information. Thoughts?

I'm interested to see how you're using libddcutil, but I'd rather not clutter this bug thread. If your code is on github I'll take a look. Or you can email me at [email protected].

Regards,
Sanford

from ddcutil.

Fixed in release 1.4.1

from ddcutil.