You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If you run in the web-interface the MP litmus test from the library (rmem link), you can see in the Graph pane, that the 'co' edge from the initial value of 'x' to the first store of 'x' is coming out of what looks like the initial value of 'y'.
The root of the problem is that the .dot file rmem generates includes HTML-like labels with PORT attribute that includes ':'. For example: <TD PORT="w(1000:1:0)">.
neato silently ignores those ports when placing edges. So, for example this:
If you run in the web-interface the MP litmus test from the library (rmem link), you can see in the Graph pane, that the 'co' edge from the initial value of 'x' to the first store of 'x' is coming out of what looks like the initial value of 'y'.
The root of the problem is that the
.dot
file rmem generates includes HTML-like labels withPORT
attribute that includes ':'. For example:<TD PORT="w(1000:1:0)">
.neato silently ignores those ports when placing edges. So, for example this:
is actually seen by neato as this:
Hence, the 'co' edge goes out from the big node that is wrapped around both values of 'x' and 'y', instead of going out of 'x'.
The tikz backend (src_top/tikz.ml) had a similar problem. This is why it uses
pp_tikz_pretty_ioid
, instead ofPp.pp_prettey_ioid
.The text was updated successfully, but these errors were encountered: