That is always the 64 million dollar question. I think the answer is yes, even though I had the guy that could fix it here:
1) What if a bus hits that guy?
2) Where we are is bleeding edge, at least based the total lack of response we got to the bug posting. But it is all Distro provided packages: We are not stirring in unknown bits. We would have had a case for a case, and the ability to go Crit-Sit, so that we would have been able to put more eyes on it.
3) Once a fix was generated via a support contract, the chances are better it would be better than our fix. By that I mean it would be Poisix compliant, and fix all the modules it needed to.
4) Once it was fixed, it would stand a better chance of making it back into the base code and helping anyone else that comes out this way in the code tree. Right now, if anyone else comes here, I have to hope they google around and find this fix. It has happened before, to be sure, but better to never have the problem at all.
A bit more on point three: When I was a VM system programmer in the 1990's, I was involved in a broken, hard down system. We patched the system up via having access to the source and got going again, and IBM took that fault back and wrote a fix for it. Our fix hit one line, an assembly language mask instruction. We did somethong to the mask, but I don't recall what.
When the final fix came out it hit a dozen modules: This error had uncovered a problem that was all over the place. Either that or IBM fixed several problems with one point patch, which they are not supposed to do...
Dan fixed this by looking at the file ownership and using that to assume that it was OK to set the RC back to 0, rather than returning an error. We have no idea if that is Posix compliant behavior: We used the CHMOD analogy to assume it would be OK. And we have no idea if there are not several other smoking guns just like this elsewhere in the code, and we just have not hit them.
Hopefully a Support Contract would take the fix and run it back up the flagpole to the Kernel folks and make sure that this problem, and all others like it are officially repaired. As it is now, every time we have to upgrade, we'll have to retrofit this fix.
1) What if a bus hits that guy?
2) Where we are is bleeding edge, at least based the total lack of response we got to the bug posting. But it is all Distro provided packages: We are not stirring in unknown bits. We would have had a case for a case, and the ability to go Crit-Sit, so that we would have been able to put more eyes on it.
3) Once a fix was generated via a support contract, the chances are better it would be better than our fix. By that I mean it would be Poisix compliant, and fix all the modules it needed to.
4) Once it was fixed, it would stand a better chance of making it back into the base code and helping anyone else that comes out this way in the code tree. Right now, if anyone else comes here, I have to hope they google around and find this fix. It has happened before, to be sure, but better to never have the problem at all.
A bit more on point three: When I was a VM system programmer in the 1990's, I was involved in a broken, hard down system. We patched the system up via having access to the source and got going again, and IBM took that fault back and wrote a fix for it. Our fix hit one line, an assembly language mask instruction. We did somethong to the mask, but I don't recall what.
When the final fix came out it hit a dozen modules: This error had uncovered a problem that was all over the place. Either that or IBM fixed several problems with one point patch, which they are not supposed to do...
Dan fixed this by looking at the file ownership and using that to assume that it was OK to set the RC back to 0, rather than returning an error. We have no idea if that is Posix compliant behavior: We used the CHMOD analogy to assume it would be OK. And we have no idea if there are not several other smoking guns just like this elsewhere in the code, and we just have not hit them.
Hopefully a Support Contract would take the fix and run it back up the flagpole to the Kernel folks and make sure that this problem, and all others like it are officially repaired. As it is now, every time we have to upgrade, we'll have to retrofit this fix.
Dan will have to stay out of the way of buses.