XPost: alt.folklore.computers
From: bowman@montana.com

On Mon, 05 Jan 2026 05:57:23 GMT, Charlie Gibbs wrote:

> On 2026-01-05, rbowman  wrote:
>
>> On Sun, 04 Jan 2026 19:41:12 GMT, Charlie Gibbs wrote:
>>
>>> On 2026-01-04, Niklas Karlsson  wrote:
>>>
>>>> On 2026-01-04, Peter Flass  wrote:
>>>>
>>>>> I look at some code and wonder "how the heck has this ever worked?",
>>>>> but the answer is that no one ever hit that combination of things
>>>>> before, or used that option.
>>>>
>>>> That's certainly the sensible explanation, but I've had scenarios
>>>> like that, even with my own code from the past, where I could swear
>>>> up and down that I myself had successfully used that code in the
>>>> exact scenario that would obviously break.
>>>
>>> Yup.  Sounds like a Schrodinbug.  It should have never worked, but it
>>> does until you look at it - and then it never works again.
>>
>> Conversely, it fails until you log a debug statement to see what's
>> going on and it works.  I'd never, never just leave the debug in place,
>> no siree.
>
> Unless the customer is screaming for a fix RIGHT NOW.
>
> But I'd go back and try to track it down once he's pacified.

It's been a day or three but I think I did. iirc it also had the charming
feature of only manifesting in the Windows build, not in Linux where I had
valgrind and electric fence.

Another mystery is why memory debuggers on Windows are expensive and
barely usable. We had a Purify license but configuring the instrumentation
was such a hassle it was rarely used. When the license came up for renewal
nobody spoke up to keep it. BoundsChecker reportedly is even worse.

--- SoupGate-Win32 v1.05
 * Origin: you cannot sedate... all the things you hate (1:229/2)