dive merge seems broken in latest master
dirk at hohndel.org
Mon Oct 15 03:28:41 PDT 2018
> On Oct 15, 2018, at 6:25 AM, Berthold Stoeger <bstoeger at mail.tuwien.ac.at> wrote:
> On Monday, 15 October 2018 11:42:26 CEST Jan Mulder wrote:
>> On 10/15/18 9:10 AM, Jan Mulder wrote:
>>> On 10/15/18 9:06 AM, Berthold Stoeger wrote:
>>>> On Monday, 15 October 2018 08:53:16 CEST Jan Mulder wrote:
>>>>> Tried to reproduce things. As I do not have 2 DCs, I fake a second one
>>>>> just by editing the ssrf XML. I cannot reproduce it (in this way).
>>>> I can trivially reproduce this by importing a divelog onto itself.
>>>> this should be easy to fix. I'm heading of to lecture now and will
>>>> look at
>>>> this in the later afternoon.
>>> Yes, this produces a crash indeed. Not 100% sure this the same one.
>> Tried to make sense of this. The first attempt crashed immediately, but
>> subsequent sessions did not. In the debugger, I saw that the trips dives
>> list (double linked list it seems) is corrupted. Then I had a couple of
>> runs the ended in an endless loop in the import on itself phase, and
>> finally, all works ...
>> So, to me, it seems non-deterministic caused by memory corruption
>> somewhere. Curious what the root cause here, but it does not feel "easy
>> to fix" to me.
> I'm quite sure that it is easy to fix, but I will only find time later this
> afternoon after work. Quite obviously, for example merge_weight_system_info()
> is broken. It should read as
> res->weight = a->weight;
> res->description = copy_string(a->description);
> or similar.
My brain wasn't working at all yesterday or I would have given it a better try.
But getting up before 3am, sitting in airplanes for 10 hours, then having a really
bad night's sleep and diving for nearly 5 hours all conspired to have me a long
way away from my personal best :-)
I did bisect at least the crash so that should give a good starting point where
I have a few minutes this morning, but if Berthold will look at it this afternoon,
then I'll instead look at PRs and whatever else has piled up in the last two
days (so we don't duplicate work).
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the subsurface