Boost interprocess fails to initialize on long-running Windows machines

[martin.apel@…]

This bug report is similar to ​#9767, but the effect described there is only part of the problem and the workaround described there may not always be applicable.

On Windows Boost Interprocess tries to determine the last boot time by looking at the system event log. This may fail, if the machine has been running a longer time or if it has only been suspended instead of rebooting most of the time. The effect is, that Boost crashes, if the boot time cannot be determined.

The workaround may not always be applicable, because a directory has to be hardcoded at compile time, which may not be accessible on all machines, where the resulting program is ever run.

So I think that instead of the workaround described in #9767 a real solution should be implemented for this case.

[William Clements <William.Clements@…>]

I saw a proposed solution in #11660

[martin.apel@…]

The solution looks OK to me, though I'm not a Windows expert. Anyway I still wonder, if it's actually necessary to have the interprocess folder name include the boot timestamp.