kicad-developers team mailing list archive

[jp charras <jp.charras@xxxxxxxxxx>]

Le 22/01/2018 à 23:45, Tomasz Wlostowski a écrit :
> On 22/01/18 17:21, jp charras wrote:
>> Hi Orson and Tomasz,
>>
>> This issue comes from the bug report 1744732.
>>
>> In some cases (easy to reproduce), the new connectivity algo see pads as connected, although it is
>> not the case.
>>
>> Attached a very simple case.
>>
>> If 2 pads are at the same Y coordinate, and 2 horizontal track segments at the same Y coordinate are
>> connected to these 2 pads, the 2 pads are seen as connected.
>>
>> If the Y coordinate is not common to all pads and segment ends, or if there are other segments
>> connected to horizontal segments, pads are seen as not connected as expected.
>>
>> The stable version (in GAL mode has also a strange behavior), and works in legacy mode.
>>
>> I am not very familiar with the new connectivity algo, this is the reason I ask you to have a look
>> into this issue.
>>
>> And I will be very happy if someone could add a few (or many) comments in connectivity_algo.h:
>> there is not a lot of comments in connectivity_algo.* files (Read: 0 comments) for this non trivial
>> code.
>>
>> Thanks.
>>
> 
> Hi JP,
> 
> I can reproduce it on my system, it's not the connectivity algo but the
> ratsnest EMST algorithm. I'll check this tomorrow (and promise to commit
> some comments explaining how the algorithm works!)
> 
> Cheers,
> Tom
> 

Thanks.
If the ratsnest EMST algorithm has an issue, you could have a look into minimun_spanning_tree.*
It is also a MST fast algorithm, that works fine, and i am thinking it is similar to the ratsnest
EMST algorithm, although the code differs.

-- 
Jean-Pierre CHARRAS