This turned out to be a very deep problem. After I fixed it & ran his
code, I get a crash in mema where Windows says memory has been modified
out of bounds. I am assuming this is his error, but in case not I think
you should give him a version with the latest fix to test with.
It turned out that the JE was doing inplace operations on
name =: *: % name
that is, =: V0 V1 N2. I didn't intend that! After an initial panic I
realized that this is actually pretty cool and will work most of the
time, which explains why there have been no complaints about it. Where
it failed was when V0 is not assignment-safe, for example is an explicit
definition. This is what Kirk had (actually, it failed only when V0 was
an adverb referring to y). Since I keep careful track of which names are
assignment-safe, the fix was easy.
When g421 failed, it left some names lying around that caused the
failure in g6x. I relaxed the test in g421 to fix both problems.
hhr
---
This email has been checked for viruses by AVG.
https://www.avg.com
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
