hash p write pathname

[masinter]

• #P"pathname" reads in as pathname
• #P"pathname" used for printing pathnames
  do reading first

[masinter]

I closed #1611 since it was included in this PR, and you will need both. This isn't changing the implementation, just adding the #P reader macro.

[nbriggs]

Presumably this doesn't affect reading the old #.(PATHNAME "foo"), since #. is just an eval reader macro and (PATHNAME "foo") is a normal way to create a pathname. There are a bunch of references to #.(PATHNAME ...) in the envos repository, though there are none that I can see in Medley or Notecards.

[masinter]

If there were any #.(PATHNAME "string") around, they would read in no problem.

[nbriggs]

I believe loading this file 'ALLPROP and remaking it would fix the duplicate COMS issue.

[nbriggs]

This seems to duplicate the RPAQQ of the CMLPATHNAMECOMS earlier in the file...?

[masinter]

The duplication of CMLPATHNAMESCOMS in this case is a leftover from Interlisp-10, where the compiled code for calling a lambda-nospread function (LAMBDA N ... (ARG N 1) ...) was different than calling a spread. It is still needed for NLAMBDA's.

[nbriggs]

OK, but I don't understand how having the exact same RPAQQ to set the CMLPATHNAMECOMS twice in different places in the file affects how lambda no-spread and nlambdas are compiled. Can you enlighten me?

[nbriggs]

And this was a newly introduced change - it was apparently OK before the addition of #P pathname, but something about that caused it to be added?

[nbriggs]

LGTM.