OLD HACKER MODE ON I was working on a multitasking implementation of DOS, and we found that the system would suddenly crash-and-hang the OS on rare occasions. Our original hint was that the sound was always on, and while we weren't ALWAYS typing, we frequently were typing.
Back in the day when programming to DOS on an IBM PC, it was well understood that the BIOS calls were safely re-entrant. Not only were they, but IBM had actually published the code for the original BIOS to allow users to rely on its behavior.
It turns out that this BIOS was not sufficiently re-entrant - it had allowed itself to be called up to two-deep, but if it was called 3 deep, that was a problem. The developer was likely sure that could not happen. But that product I was using included some very smart boards that slotted into the data bus, and indirectly it could cause its own BIOS re-entrant calls.
Also, as it turns out, if you REALLY tried hard, you could make 3 levels of BIOS calls without using an external Board, although why would you?
I wrote a lovely assembly language program that would do just that - make 3 level deep calls into the BIOS. And my Heisenbug became a bug.
Alas: when I called the OS vendor, they sort of couldn't understand what I was talking about. That code had been written by a contractor who was long-gone, and they knew they didn't understand it, and weren't going to fix it. They offered to fly me to their offices for a week to fix it (and would pay me and my employer a bounty to do so), but my employer wouldn't bite.
So, we dumped that OS and went on to another. My 3-deep tool became a standard acceptance test for evaluating OS's.
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
(no subject)
Date: 2013-10-29 06:44 pm (UTC)OLD HACKER MODE ON
I was working on a multitasking implementation of DOS, and we found that the system would suddenly crash-and-hang the OS on rare occasions. Our original hint was that the sound was always on, and while we weren't ALWAYS typing, we frequently were typing.
Back in the day when programming to DOS on an IBM PC, it was well understood that the BIOS calls were safely re-entrant. Not only were they, but IBM had actually published the code for the original BIOS to allow users to rely on its behavior.
It turns out that this BIOS was not sufficiently re-entrant - it had allowed itself to be called up to two-deep, but if it was called 3 deep, that was a problem. The developer was likely sure that could not happen. But that product I was using included some very smart boards that slotted into the data bus, and indirectly it could cause its own BIOS re-entrant calls.
Also, as it turns out, if you REALLY tried hard, you could make 3 levels of BIOS calls without using an external Board, although why would you?
I wrote a lovely assembly language program that would do just that - make 3 level deep calls into the BIOS. And my Heisenbug became a bug.
Alas: when I called the OS vendor, they sort of couldn't understand what I was talking about. That code had been written by a contractor who was long-gone, and they knew they didn't understand it, and weren't going to fix it. They offered to fly me to their offices for a week to fix it (and would pay me and my employer a bounty to do so), but my employer wouldn't bite.
So, we dumped that OS and went on to another. My 3-deep tool became a standard acceptance test for evaluating OS's.