Re: [reSIProcate] Problem using resipfaststream

["Dario Bozzali" <Dario.Bozzali@xxxxxxxxxxxxxxxxx>]

Hi all,
I noticed that in Resiprocate main SVN head it was fixed the issue in 
TlsConnection.cxx related to number of bytes returned by function SSL_read:
int bytesRead = SSL_read(mSsl,buf,count);
StackLog(<< "SSL_read returned " << bytesRead << " bytes [" << 
Data(Data::Borrow, buf, (bytesRead > 0)?(bytesRead):(0)) << "]"); 

I also noticed that the same work-around was not applied to the following lines 
(always in TlsConnection.cxx).
restBytes = SSL_read(mSsl, buffer, SSL_pending(mSsl));
StackLog(<< "SSL_read returned  " << restBytes << " bytes [" << 
Data(Data::Borrow, buffer, restBytes) << "]");
Is it not possible in this case that SSL_read() function returns -1 value?

Any news regarding the fix proposed by Justin?

I had a look at the code and in my opinion could be sufficient to change the 
following Data constructors:
- Data::Data(const char* str, int length)
- Data::Data(const unsigned char* str, int length)
- Data::Data(const char* str, int length, bool)
- Data::Data(ShareEnum se, const char* buffer, int length)
Changing mSize initialization from
   : mSize(length),
to
   : mSize(length > 0 ? length : 0),

What's your opinion regarding this change?

Thank you in advance.

Best regards,
Dario.
________________________________


-----Original Message-----
From: Justin Matthews [mailto:jmatthewsr@xxxxxxxxx] 
Sent: martedì 4 novembre 2008 18.17
To: Dario Bozzali; resiprocate-devel@xxxxxxxxxxxxxxx
Subject: RE: [reSIProcate] Problem using resipfaststream

Anyone have any objections/comments on making the changes described below to
resip::Data() constructors?

Thanks,

-justin

-----Original Message-----
From: Justin Matthews [mailto:jmatthewsr@xxxxxxxxx]
Sent: Monday, November 03, 2008 8:06 AM
To: 'Dario Bozzali'; 'resiprocate-devel@xxxxxxxxxxxxxxx'
Subject: RE: [reSIProcate] Problem using resipfaststream

Looks like the StackLog entry should be checking the return value from 
SSL_read. 

Also, any resip::Data constructors that take an int (or any functions that set 
Data::mSize) should be checking to see if the int is < 0 before assigning mSize 
(unsigned). 

You can comment out the StackLog entry for now.  Send any other errors that you 
encounter.

Thanks,

-justin

-----Original Message-----
From: resiprocate-devel-bounces@xxxxxxxxxxxxxxx
[mailto:resiprocate-devel-bounces@xxxxxxxxxxxxxxx] On Behalf Of Dario Bozzali
Sent: Monday, November 03, 2008 6:48 AM
To: resiprocate-devel@xxxxxxxxxxxxxxx
Subject: [reSIProcate] Problem using resipfaststream

Hi all,
I encountered an issue using Resiprocate 1.4 (SVN head) using resipfaststreams 
(define RESIP_USE_STL_STREAMS is commented).
I obtained an access violation (using TLS, but I think that this is not
important) I suppose because the read buffer size was -1.
In TlsConnection.cxx there are the following two lines:
        int bytesRead = SSL_read(mSsl,buf,count);
        StackLog(<< "SSL_read returned " << bytesRead << " bytes [" << 
Data(Data::Borrow, buf, bytesRead) << "]"); Executing my application SSL_read() 
returns -1.
I suppose that the access violation is caused by operator << applied to 
Data(Data::Borrow, buf, bytesRead), where bytesRead is -1.
I had a look at "size_t DataBuffer::readbuf(char *buf, size_t count)"
and "size_t DataBuffer::writebuf(const char *str, size_t count)" in 
Resiprocate\rutil\DataStream.cxx that are used when RESIP_USE_STL_STREAMS is 
not defined.
The type of size_t is unsigned int but in readbuf() and writebuf() is executed 
the following check:
[...]
   if (count <= 0)
   {
      return 0;
   }
[...]
Actually if buffer size is -1, then count is 4294967295, so the method doesn't 
exit and the writebuf() method causes access violation.
Is that right or did I make a mistake?

Thank you in advance.

Best regards,
Dario Bozzali.
_______________________________________________
resiprocate-devel mailing list
resiprocate-devel@xxxxxxxxxxxxxxx
https://list.resiprocate.org/mailman/listinfo/resiprocate-devel