<cr1901_modern>
(And by extension, if frame.f_lasti _was_ a CALL insn, why wouldn't you find at least a STORE_NAME, STORE_ATTR, etc while going up the stack. It feels like you should _always_ be able to extract a var name)
<sb0>
cr1901_modern, DUP_TOP is for handling things like self.x = x = Signal()
<cr1901_modern>
I suppose I should dump the bytecode of statements like that and see what's actually generated
<cr1901_modern>
Oh interesting... get_var_name isn't actually used anywhere within migen itself
<cr1901_modern>
erm, scratch that
<cr1901_modern>
sb0: I don't see how get_obj_var_name() is called as a part of, for example, "self.x = x = Signal()"
<cr1901_modern>
unless github's search is missing some results for "get_obj_var_name"
<sb0>
cr1901_modern, is it not? try adding a print() there and creating a Signal
<cr1901_modern>
I see, get_var_name is called in trace_back()
<cr1901_modern>
sb0: I was hoping to have the 3.6 fix done tonight, but I need to look into it a bit more.
<cr1901_modern>
Support Python 3.5 CALL_FUNCTION_KW/CALL_FUNCTION_KW_ARGs requires that we detect/ignore a few more opcodes (like BUILD_TUPLE). But I haven't yet been able to generate an example that crashes
<cr1901_modern>
Supporting*. For that reason, I think the fix should be two phase- complete Python 3.5, then refactor to support Python 3.5 and 3.6 side-by-side. Should I open a separate issue for the former?
<sb0>
_florent_, I guess the phase alignment code when buffer bypass is used can be added as external modules? it seems, this phase alignment is something you can do after the regular reset sequence has been completed?
<GitHub155>
[artiq] gkasprow commented on issue #854: @sbourdeauducq I found little issue with the bits I published. The nibbles on the Rx side were swapped and could cause confusion. I fixed it and now it's fine. The files are in the Dropbox directory I shared.... https://github.com/m-labs/artiq/issues/854#issuecomment-351219156
<GitHub173>
[artiq] gkasprow commented on issue #854: @sbourdeauducq I found little issue with the design and bits I published. The nibbles on the Rx side were swapped and could cause confusion. I fixed it and now it's fine. The files are in the Dropbox directory I shared.... https://github.com/m-labs/artiq/issues/854#issuecomment-351219156