[MLton-devel] Problem with -exn-history true

[Stephen Weeks]

> > Isn't it true that all kinds of passes assume that there are no unrechable
> > blocks? 
> 
> Nope.  There are almost never unreachable blocks, because the shrinker
> works very well.  But useless (as we just learned) and the backend are
> the only ones that depend on it, unless there are more such bugs.

And indeed, I stand corrected.  I just scanned through all of the SSA
optimizer passes, and the following three passes seem to have the same
flaw, doing analysis based on the control flow graph but then visiting
all blocks: constant-propagation, local-ref, redundant-tests.

Constant propagation is broken in exactly the same way useless is --
the diagnostics make this very obvious because they print every
variable.  I have applied the same fix there that I applied to
useless. 

Local ref looks broken.  Matthew, can you please verify and if so
apply the same fix?

Redundant tests is OK, because the properties assigned have reasonable
default values.


-------------------------------------------------------
This sf.net email is sponsored by: OSDN - Tired of that same old
cell phone?  Get a new here for FREE!
https://www.inphonic.com/r.asp?r=sourceforge1&refcode1=vs3390
_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel