Can magic show LEF obstructions ?

Yes, it can and does. There should be separate layers for each metal and obstruction for that metal (e.g., "m1" and "obsm1"). I guess that is where the error is coming from? You can see that with "see no * ; see m1,obsm1". If the obstruction area is large but is covering an area that just has routes in it, then you can end up with a false positive error from running DRC on the abstract view. Ultimately, to get the correct DRC count, you want to run DRC on the actual cell contents, not the abstract view (except that some cells like the I/O pads and SRAM core need to remain abstract because they contain geometry that magic can't interpret correctly).

yeah, this is with SRAM so hence abstract view. I didn't think it was a different layer, I do

see no *

to clear up the view and of course I wasn't seeing anything duh.

@tnt: This is one of those situations where I don't have a good feel for what is the best course of action. Ultimately I think it needs a carefully-designed abstract view for the SRAM (and the I/O) that is not taken directly from the LEF, so that it is accurate for the metal areas and possibly the transistor areas as well, without having all sorts of DRC errors in it.

Thing is ... I have no idea why the router would put something there. It is aware of the obstructions (AFAIK) and there is tons of space in that area (it's literally a trace in the middle of nowhere), so why did it put it close to the obstruction to begin with.

What's being reported as a violation is a "wide spacing rule"---a large piece of metal requires a greater spacing distance to other metal than the normal spacing rule for that metal. So if there is not a real large chunk of metal there, but only a large obstruction area, then the router is correct, and magic is flagging a false positive. It has to flag a false positive, because if the only information it has is that there is an obstruction area over there, then it must assume the worst case for checking distances from any metal geometry to that area. The worst case is that the obstruction area really is some giant chunk of metal.

Oh, I see.