Hi Klaus and all,
it appears that option parsing for the --log option is still broken, as described in this thread: https://www.vdr-portal.de/forum/index.php?thread/108924-gel%C3%B6st-logging-...
Short version: vdr stops parsing the remainder of the command line at the point of --log=2.6.
Long version:
The problem is in the getopt code in vdr.c:
case 'l': { char *p = strchr(optarg, '.'); if (p) *p = 0; if (isnumber(optarg)) { int l = atoi(optarg); if (0 <= l && l <= 3) { SysLogLevel = l; if (!p) break; if (isnumber(p + 1)) { int l = atoi(p + 1); if (0 <= l && l <= 7) { int targets[] = { LOG_LOCAL0, LOG_LOCAL1, LOG_LOCAL2, LOG_LOCAL3, LOG_LOCAL4, LOG_LOCAL5, LOG_LOCAL6, LOG_LOCAL7 }; SysLogTarget = targets[l]; break; } } } } if (p) *p = '.'; fprintf(stderr, "vdr: invalid log level: %s\n", optarg); return 2; }
Initially, the point is replaced by a NULL byte in order to be able to atoi() both numbers (if (p) *p = 0;). This replacing takes place _in place_. In theory, the replace should be reverted at the end of the code snippet (if (p) *p = '.';) BUT if one of the branches involving a break is executed, the revert is never invoked, and the original command line remains broken so that the remaining command line options are not getting evaluated. Consequently, what used to be
vdr -w 60 -g -u vdr -l 2.6 -P plugin1 -P plugin2
becomes
vdr -w 60 -g -u vdr -l 2 6 -P plugin1 -P plugin2
and the modules will not get loaded. The thread mentioned above speaks of upstart & co but I have the same issue in a docker container with explicit program start in a shell script whithout any upstart or anything. I now helped myself putting the -l at the end of the command line but I still cannot send the vdr log to a different facility because the dangling 6 is not getting evaluated.
My suggestion would be to not evaluate the optarg string directly but create a copy of optarg first, and work with the copy. This way, the original command line is unchanged. Sample patch attached, untested but it should do the trick.