Eggheads: core dump on solaris eggdrop, solution included

[Peter Johansson]

I grabbed the Eggdrop v1.6 nightly CVS snapshot from
http://www.eggheads.org/devel/ yesterday (6/20/05).

It compiled fine and is running without any problem.

I posted this on egghelp forum in beginning of this year, but was 
recently told developers don't follow that forum, so to be safe I 
reposted it here.  

Apparently the information got to you anyway before that. :-)

I didn't think of checking out latest CVS version (since I didn't
really know about it) to see if it was already fixed.

Thanks,
Peter

On 6/20/05, Bryan Drewery <lordares at shatow.net> wrote:
> This should be fixed in CVS, a patch about this was committed some months
> ago.
> 
> Give the CVS build a try and get back to us :)
> 
> Thanks.
> 
> ------
> Bryan Drewery
> 
> 
> On Mon, 20 Jun 2005, Peter Johansson wrote:
> 
> > Hi!
> >
> > I have downloaded eggdrop 1.6.17 from ftp.eggheads.org and compiled it
> > under Solaris 2.8, 2.9 and 2.10 (SunOS 5.8, 5.9 and 5.10) with gcc
> > 3.4.3. The same problem shows up on all OS versions. The problem is
> > that the bot dumps core when trying to share user files shortly after
> > trying to botlink to the hub bot.
> >
> > By running gdb on the core file I found that the segmentation
> > violation occurs in make_rand_str() which gets an erroneous string
> > length (way too large). This happens when the bot generates a random
> > password for the first time when it's linking to hub. It all boils
> > down to randint() not working properly.
> >
> > I have verified that you get results out of range from randint() by
> > using the tcl command rand.
> >
> > The problem is that RAND_MAX is 32767 which is max value that rand()
> > returns, but random() is used and on Solaris random() will return an
> > integer in the range 0 to 2**31 - 1.
> >
> > By modifying the definition of randint() macro in main.h, changing
> > RAND_MAX to 2147483647 (=2**31 - 1) it works as intended and I get no
> > core dumps. Specifically the change is to change the row
> >
> > #define randint(n) (unsigned long) (random() / (RAND_MAX + 1.0) * ((n)
> > < 0 ? (-(n)) : (n)))
> >
> > into
> >
> > #define randint(n) (unsigned long) (random() / (2147483647 + 1.0) *
> > ((n) < 0 ? (-(n)) : (n)))
> >
> > This problem can be solved by either including a check for Solaris
> > when using random() and not use RAND_MAX or by simply using rand() all
> > the time, or by
> > masking the result from random() & RAND_MAX. I'm no portability expert so
> > I'm sure you will find a good solution according to your taste. If anything is
> > unclear about the problem description, don't be shy to ask.
> >
> > This problem also affects all other functions using randint or tcl command rand,
> > which could be seen by for example auto-voice with random delay wasn't working
> > (it never voiced, probably due to waiting several hundred thousands of seconds,
> > instead of e.g. between 1 and 10 seconds).
> >
> > Please let me know if I misunderstood anything, and thanks for writing and
> > supporting the eggdrop software.
> >
> > Best regards,
> > dioid at EFNet
> >
>