1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
|
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.]
70. Debug flags doesn't take hex values as input. The manual page
can be read as suggesting it does. Might as well change it to allow
this.
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!)
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.
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.
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.
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.]
93. Check the diald archives. Time for a new archivist?
(Mailed jeremy asking if he wants to pass the torch.)
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).
|