interesting little bug...

Linus Torvalds torvalds at linux-foundation.org
Thu Mar 28 10:26:04 PDT 2013


On Wed, Mar 27, 2013 at 7:10 PM, Linus Torvalds
<torvalds at linux-foundation.org> wrote:
>
> I say "kind of" because we end up *showing* the result incorrectly. As
> far as I can tell, we now merge those two dives correctly, including
> renaming the sensor numbers on the Uemis, and all the sensor data is
> there when we save it. But it doesn't *show* the two different sensors
> correctly for the Uemis after the merge.

Indeed. I just re-sent the patch along with a fix for the but that
caused the gas mixes in the profile to be shown incorrectly.

NOTE NOTE NOTE! With the three patches, we now show the Uemis pressure
information correctly for the two different sensors, but the Petrel
profile still looks bogus. The Petrel dive has no cylinder pressure
data, but there's still some bug in the profile code that gets
confused about the gas changes when it wants to do the *interpolation*
of the known starting and ending pressures.

However, that is (once again) unrelated to the actual merging code, I
can reproduce the same bug by just taking the Petrel download on its
own and filling in the start/end pressures for the cylinders. So the
reason it shows up after the merge is that merging the Petrel and
Uemis data means that now the Petrel gets cylinder pressure data (even
if it doesn't get the *sample* pressure data, it picks up the
beginning/end pressures for the samples from the fact that the Uemis
filled them in).

Anyway, the three patches seem to improve things a lot wrt cylinder
merging. Of course, I only have one actual test-case, but it really is
a pretty nasty one (with the Petrel having insane data and listing
eight cylinders even if only two were used).

                  Linus


More information about the subsurface mailing list