Joey Hess <[EMAIL PROTECTED]> writes:
> Also, a thought:
>
> Since the bug seems to require not just a simple mremap but also support
> for relocating the pool, which I assume involves relocating pointers into
> the pool, maybe it's also time to consider some heuristic workarounds?
On each update record the size needed.
On the next update allocate a savely larger map.
If the map runs out of room give a warning, record size and
restart. If it fails more than N times abort with an error.
The size of the map is loosly related to the number of packages
available, right? So it should be possibly to guess a good upper
limit and detect runaway map size.
MfG
Goswin
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
