argv@garp.mit.edu (Dan Heller) (09/27/87)
In article <968@julian.UWO.CDN> reggers@julian.UWO.CDN (Reg Quinton) writes: >a) I have > alias pm "Peter Marshall <pm@uwovax.UWO.CDN>" > mush exands that as > Peter, Marshall, <pm@uwovax.UWO.CDN> Aliases in mush are supposed to be legal mail addresses that your mail delivery system can understand. From my experiences with sendmail (and to a lesser degree with other mailers), the address scheme of the kind mentioned above doesn't always work. So, I've restricted the form of aliases to be only addresses; names (comments) are not allowed. Some mailers will convert whitespaces to .'s unless preceded by a comma. If someone can tell me how to get sendmail to understand these addresses (without using -t), then I will consider using that approach. I'm sure it's possible and I'm just missing some aspect of the features of sendmail or other mail delivery systems. >b) I have a mail message in my box saying > From 105_172@uwo-hobbit.UWO.CDN Tue Sep 22 13:45:43 1987 > From: (Cindy Munro) 105_172@uwo-hobbit.UWO.CDN > To: (Reg Quinton) reggers@julian.UWO.CDN > Subject: Re: Discussion of Strike Issues > > When I try to reply I get > > To: reggers It seems the routine in error is the "replyall" (or, the R command). This is the error I got (in a form) when I tried to simulate this error by hand. Unfortunately, it seems that the replyall command may not work given certain situations of the format of the original message. The fix for this type of error will be made available when the patches for the current bug list is posted. >c) Several times when I mail messages I get a /bin/sh gripe. > > sh: syntax error at line 1 'end of file' unexpected. > >We've been very careful to make sure our messages conform to RFC822, I think >your address parsing isn't right and is causing these several bugs. Not that >you should feel terribly bad -- ucb/Mail has similar problems. The problem here could be related to the problem mentioned in a) above. That is, if you give sendmail a bogus address line, it'll barf and you'll get errors like the one mentioned here (based on my own experimentation). Granted, Mush should do whatever it takes to make sure sendmail gets something valid, but until that happens, I imagine that this error can be elimiated by conforming to the rules stated in a) above. Yes, RFC822 says that certain addressing schemes are legitimate, but even sendmail doesn't conform 100%. It's difficult to say exactly what is going on here since this bug report doesn't mention what the To: line contained when when he got this error, but chances are it contained something that caused his mail delivery system to barf. My mistake here is that I am using "popen" to execute the mail delivery system rather using a form of exec(). Advantages include the fact that a whole shell doesn't need to be invoked, I can trap the output and error output to the same place or separate places and have complete control and I can maintain the size of the input the command should get. This may not be fixed in the first round of patches. This fix will also enable the user to pipe messages to unix commands. Having control over the stdin, stdout, and stderr of the child of a pipe makes piping to and from non-mush (unix) commands more easily executed and error recovery more elegant.