Mailing List ArchiveSupport open source code!
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]Re: kill(2)
- To: tlug@example.com
- Subject: Re: kill(2)
- From: SL Baur <steve@example.com>
- Date: 29 Jun 2001 15:15:11 +0900
- Content-Type: text/plain; charset=US-ASCII
- In-Reply-To: "SN_Diamond"'s message of "Fri, 29 Jun 2001 13:28:53 +0900"
- Mail-Copies-To: never
- References: <3B3AF05A.32388.001@example.com> <m24rt0aqg5.fsf@example.com> <00a601c10054$070d0070$0f01a8c0@example.com>
- Reply-To: tlug@example.com
- Resent-From: tlug@example.com
- Resent-Message-ID: <9BzYJB.A.NZH.KxBP7@example.com>
- Resent-Sender: tlug-request@example.com
SN Diamond <Norman.Diamond@example.com> writes in tlug@example.com: > I nominate that for "worst abuse of the rules." :-) > (Abuse of what rules? It doesn't matter. ANY rules.) Oh no. It's not even close ... The story I am about to write is a true story. The names have been changed to protect the guilty. Back when I working for a major DOD contractor, they were simultaneously going from using mostly VAX/VMS & Fortran to Unix (HP/UX as it turns out) & Ada. At this time, there was severe lack of experienced Unix people and of course we were all put into rush courses to learn Ada. One particular project was experiencing some difficulty with a demo they were putting together. This was the first time any of the programmers had used either Ada, or the brand new Motif UI toolkit. They decided to make every screen widget a separate Ada task (interrupt reentrancy, what's that?). They also decided to make this program setuid root. Among other bugs I had to help them fix, they were an Ada function called DELETE that was implemented with the unlink(2) system call. They were calling this function on a directory. Since the program was setuid root, the kernel was happy to do the unlink. But then they were wondering why disk space kept disappearing and why they constantly had to reboot to recover it. Not understanding concurrency and the concept of interrupt reentrancy, coupled with the fact that Motif is not interrupt reentract, they failed to lock any of the graphics calls. This produced seemingly random results. By the time I was called in, they had mostly ruled out hardware failure since no matter what machine they tried, it didn't work. The leading candidate was sunspots affecting the computer memory. I was told later that the reason why they did it this way was for efficiency. Whatever. Anyway, there it is. A $20,000,000 contract dependent upon a graphics program with maybe the most ridiculous architecture of all time, Ada tasks are relatively heavy weight, using Motif and setuid root. This is my candidate for "worst abuse of the rules".
- References:
- Re: kill(2)
- From: vp@example.com (Viktor Pavlenko)
- Re: kill(2)
- From: SL Baur <steve@example.com>
- Re: kill(2)
- From: "SN_Diamond" <Norman.Diamond@example.com>
Home | Main Index | Thread Index
- Prev by Date: Re: Q.
- Next by Date: Re: Q.
- Prev by thread: Re: kill(2)
- Next by thread: Re: kill(2)
- Index(es):
Home Page Mailing List Linux and Japan TLUG Members Links