intermittent flaw in gnumake deps on Win32 ..

[Norbert Thiebaud]

On Fri, Mar 30, 2012 at 6:56 AM, Christian Lohmaier
<lohmaier+libreoffice at googlemail.com> wrote:
> On Fri, Mar 30, 2012 at 1:05 PM, Matúš Kukan <matus.kukan at gmail.com> wrote:
>> On 30 March 2012 11:36, Michael Meeks <michael.meeks at suse.com> wrote:
>>>        Regina's problem comes down to a corrupt .d file.
>>> [...]
>>>        One more data-point, that file was exactly 4096 bytes large:
>>
>> What does that mean ?
>
> Well - it is either a coincidence, or not.. A multiple of 1024 i.e
> exactly 4KB in size..

and 4K is the typical memory page, and the kidn of unit you would see
for partial write. the final write to device.
this behavior can be seen if you have a running process that write to
a log and look at the log file while it is running
the log will grow by 4K increment until you ffllush() of fclose() the file.

So that what one would expect to see if the filter-showincludes.pl was
interrupted abruptly.

Norbert