The next bug report will receive number: 89. 48. I should get someone with an alpha to test diald out in that environment so I can get rid of type size assumptions. Before I do that I need to look through the code for any places where I use "int" or "long" and really mean u_int32_t. (Mostly this will be with IP addresses.) Simiarly, I should look for uses of short where I really want u_int16_t, or char where I want u_int8_t. [I have one report that indicates diald won't even compile with the glibc implementation being used on alphas. I suppose I'll have to grab glibc to fix this myself. I hope I can make it consistent between the two libc's. Yuck. (Check out Philip Blundels patches...)] 59. By the way, ip-up options inside a restrict does not seem to work. The last ip-up script is always chosen, no matter the time. [As designed, but this would be neat to be able to do...] 62. Diald can get "stuck" writing out to a monitor line if the other end does something stupid in the middle of a write on our part. [FIXED? Race condition possible with writes due to alarms could cause the data structure listing monitors to get corrupted. Changed alarms to not use an interupt. Race conditions all gone. Still something funny about the opening code for the FIFO.] 64. If the load goes so high that diald can't get everything out of the slip connection, odd stuff starts to happen. [This requires some major rewrites. A start was getting rid of alarms() to do timers. Next step is dealing with monitor writing on overloaded systems. ] 65. Something fishy in the impluse rules stuff. [FIXED? I rewrote some stuff and applied some patches. Let's see what the testers say.] 71. Using scanf to do parsing is just begging for trouble. I should rewrite the parsing so it doesn't use scanf at all. I should also go over the diald code to make sure that it can't suffer from buffer overruns causing it to run arbitrary code or just crash. [Made some starts on this. Still to much dependence on sscanf for my taste.] 72. [Giuseppe Vacanti <gv@deselby.xs4all.nl>] Should diald perhaps be putting the fifo into /var/run by default? It is a run time entity after all. (Precident from gpm!) [ Mike: No. /var/run is for per-boot things. The diald fifo is persistent across boots and may be carrying permissions settings that the user wants to keep (Mine have a particular group for instance. ] 74. [Roger C. Pao] If I run diald with no arguments and no /etc/diald.conf, it will core dump. add_device() is never called before check_setup(). [FIXED.] If I run diald with "ttyS1" as an argument, it does not core dump, but it's not finding the device. I have to use "/dev/ttyS1" as the device either on the command line, or in the /etc/diald.conf file. [This should be noted in the manual pages.] 75. It might be worth mentioning that you should prefer to use /dev/ttyS more strongly in the docs. Perhaps a copy of tytso's message about the subject, and a description of the setserial flags that can bugger you up. (Maybe a clearer description of what the pppd patches do as well.) 76. [Jamie Penner] Seems sometimes masquerading combined with firewalling can keep diald from seeing any packets. There may be a problem here that needs to be looked into but it is not clear. [I suspect this could happen if you want to try keep packets not marked as "from us" going out on the link, since the firewall rules could take them out before the masquerading gets a look at them. I should check with Jamie if this was his configuration.] 77. It should be made clearer in the documentation that setting local and remote does not set pppd's idea of local and remote in any way and that "pppd-options X:Y" is sometimes necessary. 78. Diald should accecpt a "-v" (version) flag on the command line. (Hmm. Verbose mode sort of does this...) 79. See message dna/1997/13/518 for a good suggestion to be added to the FAQ. 80. See message dna/1997/13/699. It describes a problem with diald connecting to incoming calls that certify via chap/pap without any login dialog. 82. Make note about slackware brain damage in FAQ. See message 1997/12/854 for more information. 83. See message 1997/12/474 from Jim DiFronzo for some patches to dev.c. [These patches aren't quite right. I've asked for an update with some suggestions.] 85. Add code to locking to check if own process id is stored in the lock file. [ Mike: Done - treat as a stale lock file and go round again. ] 86. Look at Arrigo Triulzi's message about named configuration. Probably need him to rework it a bit for inclusion into the general FAQ. 88. Diald does not report the forcing/blocked state on the monitor line. It should do so. [ Mike: It does now if the monitor requests the new protocol. ] 89. Fix dctrl so that the window size doesn't jiggle. Also document the recent changes to command line parameters that came from Kurt. [Wiggle should be fixed. Window is slightly wider now.] 90. Add in something like "sub-restrict" that could cut a hole out of a restrict time period. This makes restrict statements more like: "restrict, add-restrict, sub-restrict". 91. The connect script timeout doesn't work. All it succeeds in doing is cancelling the shell. In specific cases it might be possible to use exec chat, rather than just chat, and it is possible that bash (I use ash on a minimal system) spots the tail recursion in the simple cases. However, I think the general case requires that a new process group be created (with precautions against input being stalled because the terminal is associated with the wrong group, and a process group kill should be used. [ Mike: Done. ] 92. The get_pty routine drops through if it doesn't find one. I actually had the machine configured with only a few new stle PTYs, and none in the pty[pqrs] range. The result is various strange error messages. It might be an idea to include the expected range in the FILES section of the man page. [Partially fixed. No reports a sensible error if no pty is found. It should still be rewritten to look at all possible names.] 94. Dctrl man page is wrong with regard to -control-window behavior. Iconifying does not prevent an icon from appearing in this case. 95. Document how to use a connect script to dial more than one number. 96. Point to diald-top and the new diald monitoring deamon in the docs. Might as well point to any other contrib stuff promeniently as possible, maybe to config tools as well. 97. See message dna/1997/17/686 for a start on patches to rewrite addresses when diald is going up. It's not a total solution, but it's a nice start. Best would be to make masquerading do the right thing (TM).