kicad-developers team mailing list archive

[Wayne Stambaugh <stambaughw@xxxxxxxxx>]

On 9/21/2016 9:17 PM, Cirilo Bernardo wrote:
> On Thu, Sep 22, 2016 at 10:07 AM, Wayne Stambaugh <stambaughw@xxxxxxxxx> wrote:
>> On 9/21/2016 7:27 PM, Cirilo Bernardo wrote:
>>> OK, I've updated the branch with the following changes:
>>>
>>> 1. removed wxT() from kicad2step and dialogs. The remaining wxT()
>>> instances are created by wxFormBuilder.
>>>
>>> 2. refined the Export STEP GUI for cases in which the exporter
>>> fails (returns an error or segfaults).
>>
>> Thanks for the fixes.  I got everything merged on my computer at work so
>> I'll just pick up where I left off tomorrow.
>>
> 
> No problem; thanks for testing the branch.

I tested this and everything looked fine.  I did notice in your last
commit that you used stderr.  I'm not sure that has any meaning on
windows but it didn't seem to break anything.  I committed you kicad
step export branch to the product repo.  Great job.  Thanks.

I do have one comment.  The vrml exporter includes the silk screen and
traces which results in a much more detailed board model.  It would be
nice if step export had this as well.

> 
>>>
>>> It also just occurred to me that sometimes the OCE library may
>>> cause a hang. I can work on a generic dialog to launch an
>>> external app which connects to the apps stdout + stderr and
>>> which has a CANCEL button to kill the process - any comments?
>>> Should I put such a dialog into the "common" library?
>>
>> I'm wondering if you could make something abstract enough to work in all
>> the places where we run external apps.  These dialogs always seem pretty
>> task specific like the netlist and bom dialogs?  You could try I guess.
>>
> 
> The best abstraction I can think of at the moment is to instantiate a dialog
> which simply has a window showing the stderr + stdout of the running app
> and a cancel button to kill the process if necessary. All other customizations
> should be done in a parent window. I'll come up with something and apply it
> to the BOM and netlist generation as well.

Given that we've never had any issues with the bom and netlist external
processes, it may not be worth the effort although I'm not opposed to
the idea.

> 
>>>
>>> The fact that a process using OCE can hang brings up the
>>> question of whether it is better to leave kicad2step as a
>>> separate app or whether it is generally OK as a plugin and
>>> the odd crash due to bugs in OCE and/or the STEP/IGES
>>> models would be acceptable. We can stuff the plugin
>>> invocations into their own thread and check for completion,
>>> but unlike the case with a separate process, we cannot
>>> guarantee there is no memory corruption or leakage.
>>> Any thoughts?
>>
>> Ughh!  You're not making me feel any better about oce.  It would be nice
>> to be able to kill a rouge process though.  Doesn't the oce library api
>> provide some kind of error reporting capability?
>>
> 
> OCE does have an error reporting scheme but I've seen it hang
> indefinitely while performing some operations such as shape healing
> on defective STEP models. In other instances it eventually stops
> after it has hogged the maximum memory that the system would
> give it. In both cases it's not possible to get error information from
> within OCE. Unfortunately such bad behavior is not unique to OCE; the
> MCAD world in particular seems to be full of buggy libraries, but to
> be fair the tasks involved are incredibly difficult and users no doubt
> are always coming up with cases that the developers hadn't checked for.

Are we the only project that pushes these libraries that hard?  We do
seem to find our fair share of library issues.

> 
>>>
>>> Somewhat off-topic: grep shows me that the source code
>>> and headers are full of wxT().  Since wxT() had been
>>> deprecated years ago and KiCad is no longer compatible
>>> with versions of wxWidgets which required wxT(), perhaps
>>> we should ask devs to purge wxT() from the headers and
>>> sources which they touch? I think that might also get devs
>>> into the habit of not using wxT() - even I still use it without
>>> realizing it - bad habits die hard. :)
>>
>> I caught myself using wxT() while writing the legacy schematic plugin
>> several times.  It will take time to get used to it but we will get
>> there.  If you happen to be working on a file, please feel free to
>> remove them.  I don't think we need to do the mass purge just yet.
>>
> 
> OK; I might create some patches for the bits I'd worked on in the
> past like VRML export and IDF export.

For small change sets (<= 5 commits), please send me patch sets using
format-patch or send-email.  It's easier than me having to add remotes
and go through the whole pull rebase cycle.

Cheers,

Wayne

> 
> - Cirilo
> 
>>>
>>> - Cirilo
>>>
>>> On Thu, Sep 22, 2016 at 3:04 AM, Wayne Stambaugh <stambaughw@xxxxxxxxx> wrote:
>>>> Cirilo,
>>>>
>>>> I just tested this since you fixed the windows extension issue.  The
>>>> menu item is enabled but I always get an "Unable to create step file
>>>> whenever there are spaces in the file name and/or path."  You didn't by
>>>> chance forget to double quote the command line string did you?  If you
>>>> don't, spaces in file and/or path names in command strings will fail.
>>>>
>>>> Just a couple of quick comments nothing major.  wxT() macros are no
>>>> longer required in wx3 so try to remember not to use it anymore since
>>>> it's slated to be deprecated in the future.  It's also not necessary to
>>>> convert path separators in strings when you are already using
>>>> wxFileName.  You can use wxFileName::GetFullPath() which will return the
>>>> native separators no matter what you feed it with.  You can also convert
>>>> to the unix file separator for storage by using wxFileName::GetFullPath(
>>>> wxPATH_UNIX ).  This removes the need for #ifdef WINDOWS/#endif to do
>>>> the separator conversion.
>>>>
>>>> Cheers,
>>>>
>>>> Wayne
>>>>
>>>> On 9/19/2016 3:53 AM, Nick Østergaard wrote:
>>>>> Looks good, I will test it soon. But I noticed that it looks like you
>>>>> did not use the copyright template copyright.h from the root of the source.
>>>>>
>>>>>
>>>>> Den 19/09/2016 09.46 skrev "Cirilo Bernardo" <cirilo.bernardo@xxxxxxxxx
>>>>> <mailto:cirilo.bernardo@xxxxxxxxx>>:
>>>>>
>>>>>     The kicad-step feature branch now implements a STEP Export. The menu
>>>>>     item may need a new icon (I lazily reused the IDF icon). Any testing and
>>>>>     comments would be appreciated. The kicad2step utility which performs
>>>>>     the conversion is of course dependent on OCE and is only built when
>>>>>     KICAD_USE_OCE is defined. The "Export STEP" menu item is disabled
>>>>>     if the kicad2step executable is not found in the same directory as the
>>>>>     pcbnew executable.
>>>>>
>>>>>     https://code.launchpad.net/~cirilo-bernardo/kicad/+git/kicad-oce/+ref/kicad-step
>>>>>     <https://code.launchpad.net/~cirilo-bernardo/kicad/+git/kicad-oce/+ref/kicad-step>
>>>>>
>>>>>     - Cirilo
>>>>>
>>>>>
>>>>>     _______________________________________________
>>>>>     Mailing list: https://launchpad.net/~kicad-developers
>>>>>     <https://launchpad.net/~kicad-developers>
>>>>>     Post to     : kicad-developers@xxxxxxxxxxxxxxxxxxx
>>>>>     <mailto:kicad-developers@xxxxxxxxxxxxxxxxxxx>
>>>>>     Unsubscribe : https://launchpad.net/~kicad-developers
>>>>>     <https://launchpad.net/~kicad-developers>
>>>>>     More help   : https://help.launchpad.net/ListHelp
>>>>>     <https://help.launchpad.net/ListHelp>
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Mailing list: https://launchpad.net/~kicad-developers
>>>>> Post to     : kicad-developers@xxxxxxxxxxxxxxxxxxx
>>>>> Unsubscribe : https://launchpad.net/~kicad-developers
>>>>> More help   : https://help.launchpad.net/ListHelp
>>>>>
>>>>
>>>> _______________________________________________
>>>> Mailing list: https://launchpad.net/~kicad-developers
>>>> Post to     : kicad-developers@xxxxxxxxxxxxxxxxxxx
>>>> Unsubscribe : https://launchpad.net/~kicad-developers
>>>> More help   : https://help.launchpad.net/ListHelp
>>