gEDA-dev: Bug reports for gattrib

[Gabriel Paubert]

On Sat, Mar 31, 2007 at 05:47:38PM -0400, Stuart Brorson wrote:
> >Where is the bug tracker (if any) for the gattrib program? I find
> >that if an object has multiple attributes with the same name, the
> >gattrib program will make a mess and not write the attributes out
> >correctly.
> 
> As somebody else pointed out, put it in the main gEDA/gaf bug tracker
> on SF.
> 
> At some point in the processing, Gattrib forms a list of name/value
> pairs, using the attrib name as the key.  This is done while it is
> figuring out what the column names should be.  That's probably why
> things are getting munged.
> 
> However, my question is:  Why are you attaching attributes to the same
> component which have the same name?   Is it something like
> vendor/vendor part no. for different vendors?  And what should the
> correct behavior be?
> 

In my case, I've done it to override net= for logic chips. It was in a
board which controls GaAs FET based microwave switches like Hittite's
HMC270. I have a translation layer from standard TTL levels to 0/-5V
followed by a '541 (actually a CMOS version) buffer . On the '541
I have net=GND:20 and net=-5V:10.

I just checked this schematic under gattrib, and it shows only one
net= for the chip (I expected it). When saving the result, I had two
copies of net=GND:20.

I don't know what the correct solution is: maybe gattrib should ignore 
net attributes (they are really a bit different from most others, since
they can be partially overridden) or give up (perhaps with a warning) on 
an attribute with multiple instances and admit that it is outside its 
domain of competence. 

	Regards,
	Gabriel