There's not much more to show short of the example from the documentation. $^M isn't for general control over use of memory; it's a specific place you can pre-allocate a buffer to be used formatting error messages in the case that your process hits a memory limit (and hence couldn't request more memory from the OS at that point in time). Think of it like pulling a sheet of paper out of the ream and setting it in a safe place next to the copier so you have someplace to write a "Copier out of paper" sign when you run out.

I have problems understanding the doc.

If "running out of memory is untrappable" how can I be able to run code anyway?

Or is it still trappable via $SIG{__DIE__} ?

In that case why can't I just pre-allocate variables in the sig-handler's closure instead of requesting more memory?

Why $^M at all?

Honestly I don't know how to simulate "running out of memory" for experimenting.

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery}

There's a bit more of an explanation in perldiag for "Out of memory during request for %s" and I think this is presuming that you've compiled using perl's malloc.

    Out of memory during request for %s
        (X)(F) The malloc() function returned 0, indicating there was
        insufficient remaining memory (or virtual memory) to satisfy the
        request.

        The request was judged to be small, so the possibility to trap it
        depends on the way perl was compiled. By default it is not
        trappable. However, if compiled for this, Perl may use the contents
        of $^M as an emergency pool after die()ing with this message. In
        this case the error is trappable *once*, and the error message will
        include the line and file where the failed request happened.

So iff you've got a perl with perl's malloc and you give a buffer to $^M and then run out of memory perl will die and write its error message to that buffer. In that case yes then you could trap that one exception, and you could see where the exception occurred (as opposed to just getting an "Out of memory" error from the OS' malloc maybe).

As far as testing I *think* you could use something like ulimit -m to artificially lower how much memory's available to child processes. Allocate a small buffer as shown in the example, then try and create a large string (run ulimit -m 512 then try perl -E '$^M=(q{}x(1<<16));eval { $x = (q{FOO} x (512 * 1024 * 1024)); }; if($@){die $@;}')

The cake is a lie.
The cake is a lie.
The cake is a lie.

Honestly I don't know how to simulate "running out of memory" for experimenting

That is one of those things that seems like it should be quite easy...that is, until you start to think about actually doing it!
If I were going to try, I think I would start with a virtual machine with very little memory allocated and then get my program to deliberately create lots of arrays of arrays.

iirc $^M is a leftover or similar.

It worked only if perl is built to use its own malloc instead of the system one (see comments in malloc.c). UPDATE Perl's malloc is used only if -Dusemymalloc is used when building perl.

See also: Is $^M a leftover April Fool?

L*

Just in case it's of use, I keep a general list of references on Perl memory problems:

• Re: Perl Memory problem ... (Memory Tools References)

Many Thanks! the link you paste is a very useful collection!




---

---

I am trying to improve my English skills, if you see a mistake please feel free to reply or /msg me a correction

Hi,

"perl lack global variable/memory for use"

Please see MCE::Shared which works with all perls back to 5.8 and on all OSes and under threaded and non-threaded perls.

Hope this helps!