Some issues with error conditions

Some issues with error conditions

[Dmitriyi Paduchikh]

Hello,

Evaluating expression
(let ((process (start-process "false" nil "false")))
  (sleep-for 1)
  (process-send-string process "Hi there!"))

produces an error of most general kind:

Debugger entered--Lisp error: (error "Process false not running")
  process-send-string(#<process false> "Hi there!")
  (let ((process ...)) (sleep-for 1) (process-send-string process "Hi there!"))
  eval((let ((process ...)) (sleep-for 1) (process-send-string process "Hi there!")))
  eval-last-sexp-1(nil)
  eval-last-sexp(nil)
  call-interactively(eval-last-sexp)


Shouldn't it be something like process-error instead? IMHO signalling
general error in this situation makes catching of errors related with
processes more difficult.

And also end-of-file (end of file during parsing),
invalid-read-syntax, scan-error (not sure about this) probably might
be all in the category parse-error as such condition already exists.

-- 
Dmitriyi Paduchikh

Re: Some issues with error conditions

[Dmitriyi Paduchikh]

I wrote:
[...]
> And also end-of-file (end of file during parsing),
> invalid-read-syntax, scan-error (not sure about this) probably might
> be all in the category parse-error as such condition already exists.

Sorry, the condition parse-error is defined in calendar/parse-time.el,
so probably it's not appropriate to do.

-- 
Dmitriyi Paduchikh

Re: Some issues with error conditions

[Richard M. Stallman]

    Shouldn't it be something like process-error instead? IMHO signalling
    general error in this situation makes catching of errors related with
    processes more difficult.

It sounds like a good idea.  However, this is a new feature, not
in any sense fixing a problem, so we should not change it now.
I will add it to etc/TODO.

    And also end-of-file (end of file during parsing),
    invalid-read-syntax, scan-error (not sure about this) probably might
    be all in the category parse-error as such condition already exists.

I didn't know that `parse-error' was defined.
It turns out to have been defined by a specific package
for its own error conditions.

We could redefine it as a broader category.
Again, we should not do so now.  I will add it to etc/TODO.