Re: [AUCTeX] Text corruption

[David Kastrup]

"Axel E. Retif" <address@hidden> writes:

> David,
>
>
> On 12/16/2012 02:44 PM, David Kastrup wrote:
>
>> Tassilo Horn <address@hidden> writes:
>
> [...]
>
>>> That's a really strange problem.  It doesn't sound like an Emacs or
>>> AUCTeX problem but more like a colleague playing a trick on you.
>>
>> More like the computer playing a trick on him.  i (0x69) and m (0x6d)
>> differ by a single bit, bit 2.  Time to run some memory checker.
>
> I ran memtester for about an hour, without any problem.

I've had this more than once: different CPU generations and chipsets,
nothing to be seen under regular memtest.  It might be related to some
memory accesses (like opcode fetches) being affected more by timing
problems than others, or it might be partly a refresh problem that just
does not emerge in the quasi-regular patterns from a memory test.

But program errors (and/or a bad bit in program code or address) are not
likely to accidentally cause merely a single bit change.  While the
"evidence" so far is minuscule, my current guess would be bad memory.

-- 
David Kastrup