Is -w really useful?


#1

I would like to use the -w commandline option always, but any useful
output is always obscured by message from the standard libraries:

For example:

/usr/local/lib/ruby/gems/1.8/gems/rio-0.3.7/lib/rio/if/grande.rb:404:
warning: useless use of > in void context
/usr/local/lib/ruby/gems/1.8/gems/rio-0.3.7/lib/rio/if/grande.rb:580:
warning: useless use of < in void context
/usr/local/lib/ruby/1.8/tk.rb:2313: warning: redefine encoding=
/usr/local/lib/ruby/1.8/tk.rb:2316: warning: redefine encoding
/usr/local/lib/ruby/1.8/tk/font.rb:718: warning: instance variable
@compoundfont not initialized
/usr/local/lib/ruby/1.8/tk/font.rb:671: warning: instance variable
@compoundfont not initialized
/usr/local/lib/ruby/1.8/tk/font.rb:718: warning: instance variable
@compoundfont not initialized
/usr/local/lib/ruby/1.8/tk/font.rb:671: warning: instance variable
@compoundfont not initialized
/usr/local/lib/ruby/gems/1.8/gems/rio-0.3.7/lib/rio/rl/uri.rb:78:
warning: method redefined; discarding old base=
/usr/local/lib/ruby/gems/1.8/gems/rio-0.3.7/lib/rio/cp.rb:209: warning:
useless use of < in void context
/usr/local/lib/ruby/gems/1.8/gems/rio-0.3.7/lib/rio/cp.rb:309: warning:
useless use of > in void context

Should such messages not be banned?


#2

On Mar 27, 2006, at 5:29 AM, Wybo D. wrote:

I would like to use the -w commandline option always, but any
useful output is always obscured by message from the standard
libraries:

[snip warnings]

Just to be clear, Rio is not a standard library (though TK is).

Should such messages not be banned?

I do agree that standard libraries should not trigger warnings.

James Edward G. II


#3

Wybo D. wrote:

/usr/local/lib/ruby/1.8/tk.rb:2316: warning: redefine encoding
/usr/local/lib/ruby/gems/1.8/gems/rio-0.3.7/lib/rio/cp.rb:209:
warning: useless use of < in void context
/usr/local/lib/ruby/gems/1.8/gems/rio-0.3.7/lib/rio/cp.rb:309:
warning: useless use of > in void context

Should such messages not be banned?

Some warnings are really just informational. For example, redefining a
method isn’t ‘wrong,’ and it can be how your program is supposed to
function. Or having an uninitialized instance variable might not matter

  • after all, I don’t think there is a difference between that and
    setting it to nil.

However, it could be dangerous or unintentional, so you get a warning.
As with all debugging output, you have to filter through it sometimes.

-Justin


#4

On Tue, 28 Mar 2006, Justin C. wrote:

Some warnings are really just informational. For example, redefining a method
isn’t ‘wrong,’ and it can be how your program is supposed to function. Or
having an uninitialized instance variable might not matter - after all, I
don’t think there is a difference between that and setting it to nil.

However, it could be dangerous or unintentional, so you get a warning. As
with all debugging output, you have to filter through it sometimes.

Sure, that’s all true, and I want to see such messages to warn myself
for
possibly bad programming habits.
But when I see lots of such message emerge from libraries that I didn’t
write myself, I’ll easily miss the ones that really matter to me and can
be influenced by me.


#5

On Tue, 28 Mar 2006, Robert D. wrote:

Wouldn’t it be nice to have a perl/ada style

pragma :warnings_on
pragma :warnings_off
or
pragma :warnings, level

at a dynamic base.

Well, you can already say: $VERBOSE = true or $VERBOSE = false.
Isn’t that the same?


#6

On 3/27/06, Wybo D. removed_email_address@domain.invalid wrote:

[messages skipped]
don’t think there is a difference between that and setting it to nil.


Wybo

Wouldn’t it be nice to have a perl/ada style

pragma :warnings_on
pragma :warnings_off
or
pragma :warnings, level

at a dynamic base.

Now for a given (standard) library we could just say

e.g.
module MyModule
pragma :warnings 0
end

Just a thaught

Robert


Deux choses sont infinies : l’univers et la bêtise humaine ; en ce qui
concerne l’univers, je n’en ai pas acquis la certitude absolue.

  • Albert Einstein

#7

On Wed, 29 Mar 2006 07:17:10 +0900, Wybo D. removed_email_address@domain.invalid
wrote:

Well, you can already say: $VERBOSE = true or $VERBOSE = false.
Isn’t that the same?

Actually, $VERBOSE can take on three semantic values: nil, false, true.
Further, behaviour can differ slightly if -W0, -W1, -W2 (ie, nil, false,
true (and -w == -W2) are set from the command line as opposed to setting
$VERBOSE in the script (parse-time vs runtime). This means that doing
something akin to:

module Kernel
def no_warn
begin
verbose = $VERBOSE
$VERBOSE = nil
yield
ensure
$VERBOSE = verbose
end
end
end

no_warn do
puts (0…5)
end

doesn’t quench the warning, but -W0 does. Strangely, changing that
block to:

no_warn do
puts “foo”
puts (0…5)
end

results in yet different warning behaviours.

At any rate, a catchable warning system might be a good idea and
encourage people to wrap ‘warnable’ code in blocks that don’t emit
particular warnings – so something like:

no_warn :method_redefined do
# (meta)code that redefines methods
end

would explicitly turn off the method redefinition warning (presumably we
are doing that on purpose here), but still emit any other warnings the
code might generate.

Just a thought.

andrew