Re: [bug-mailutils] Stream usage

[Simon Walter]

Hello, 

I have another problem related to this.
Even when I create a copy of the message and use this copy to parse
through the message I can't use the attributes of the original message
for attribute_set_deleted().
The pointer _get_flags in the attribute struct gets invalid.

Marking the message as deleted before I work with the stream is no
problem but thats a bad solution for me, because I only wan't to
deleted a message if processing and forwarding went fine.

Creating new attributes and set them to the message also fails at the
destruction of the old attributes.

Any suggestions?

Alain Magloire <address@hidden> writes:
>> -----Original Message-----
>> From: address@hidden [mailto:bug-mailutils-
>> address@hidden On Behalf Of Simon Walter
>> Sent: Wednesday, September 07, 2005 1:27 PM
>> To: address@hidden
>> Subject: [bug-mailutils] Stream usage
>> 
>> 
>> Hello
>> 
>> I'm using mailutil libs in a small mailrobot.
>> 
>> while developing the mailbox is a simple mailfile.
>> in productive the mailbox is pop3.
>> 
>> I have written a function to search with a regular expression through
>> the mailbody and return everything after a separator.
>> 
>> when I use the mailfile I can call the function twice, but when i use
>> a pop3-account the second call of the function doesn't enter the
>> while-loop.
>> 
>> so there must be a difference between a memory-based stream and a
>> file-based...
>> is this a bug or am I doing something wrong?
>> 
>
> It's been a long time since I play with mailutils, but basically yes, there
> is a difference, not so much between the stream but the underlying
> implementation (POP3 vs File access, Mailbox).  For POP3 protocol, when you
> issue a command to retrieve data whether it is the header, the body or the
> full message, the command is not cancelable.
>
> So you may have broken out of the loop too early (before reading the entire
> message/body) because of regex match.  In this scenario, when you come back
> the second time to read another message, the stream reading returns EBUSY or
> something like that, meaning the last TOP/RETR command sent to the pop
> server is not complete.
>
> You can take a few steps to remedy:
> - do not break out of the loop before you finish reading the message
> - Or make a copy of the message in a second stream. For example use a
> memory_stream(or a file_stream) to hold temporary the message then do the
> matching on that stream.  This will isolate you form network hiccup for
> protocols like POP3 that are limited.
>
> We take some steps to hide the underlying protocol to the user, but in the
> end .. it may not be enough.  I'm thinking maybe we should look at a
> shadowing mailbox or something similar to hide those mundane
> protocol/networking detail to the users ... hmmm...
>
>> <code>
>> char *mc_get_field_value(data_t *data, regex_t *regexp, char *sep) {
>> 
>>   char *value = NULL;
>>   body_t body;
>>   stream_t stream;
>>   char *p;
>>   char buf[128];
>>   off_t off = 0;
>>   size_t rd;
>> 
>>   int regex_ret;
>>   regmatch_t regmatch[RE_SIZE];
>> 
>>   /* ||| */
>> 
>>   message_get_body(*(data->msg), &body);
>>   body_get_stream(body, &stream);
>> 
>>   while(value == NULL
>>         && stream_readline(stream, buf, sizeof(buf), off, &rd) == 0
>>         && rd != 0) {
>>     off += rd;
>> 
>>     if ((regex_ret = regexec(regexp, buf, RE_SIZE, regmatch, 0)) == 0) {
>>       if((p = strstr(buf,sep)) != NULL) {
>>         value = strdup(p+strlen(sep));
>>         if((p = strchr(value,'\n')) != NULL)
>>           *p = '\0';
>>         break;
>>       }
>>     }
>>   }
>> 
>>   return value;
>> }
>> </code>

-- 
Mit freundlichen Grüßen

Simon Walter
NETHINKS GmbH, Technik
Bahnhofstraße 16
36037 Fulda
Tel: +49 661 25000-0
Fax: +49 661 25000-49
eMail: address@hidden