Re: [PATCH] lower cap on max_cmd_len to fix hppa2.0w-hp-hpux11.00

[Robert Boehne]

Michael:

I'm not setting CONFIG_SHELL either, nothing special, just ./configure ;
make;
In short, I don't think that a string as long as a command line should
eat much memory, certainly not enough to limit it's use in general.
Perhaps you could also look at this from the standpoint that if you have
a project that uses piecewise linking under HPUX 11.0, you won't even
be able to compile it because Libtool would take up more memory than
you have free.  As apps go, Libtool isn't what I'd call a memroy hog,
and as you note it works just fine on your Solaris machines.  How much
memory is the libtool configure using?  Are there shell bugs that you
don't have patches for?  IMHO, this isn't a libtool problem.

Sorry,

Robert


Michael Elizabeth Chastain wrote:
> 
> Hi Robert,
> 
> (I'm taking the liberty of adding the public list back to the
> recepients.)
> 
> Okay, let me try to convince you ...
> 
> What CONFIG_SHELL are you using?
> 
> I am not setting CONFIG_SHELL at all, so I am getting the default of
> /bin/sh.  When I set CONFIG_SHELL=/bin/ksh, I get the same result as you:
> 
>   checking the maximum length of command line arguments... 786433
> 
> There's nothing at all in the docs about CONFIG_SHELL, let alone a
> notice about which systems need the user to set a non-default value
> for CONFIG_SHELL.
> 
> IMHO it's better that configuration work straight out of the box than
> it is to capture the performance improvement between a 390K max_cmd_len
> and a 780K max_cmd_len.  BTW, that's why I tested on a bunch of Solaris
> machines.  They are reporting 390K with the existing loop (18 iterations)
> and there is no change for them with my loop (17 iterations).
> 
> MichaelC
> 
>   From address@hidden Mon Jul  2 14:55 PDT 2001
>   Sender: address@hidden
>   Date: Mon, 02 Jul 2001 16:55:11 -0500
>   From: Robert Boehne <address@hidden>
>   Organization: Ricardo Software
>   To: Michael Elizabeth Chastain <address@hidden>
>   Subject: Re: [PATCH] lower cap on max_cmd_len to fix hppa2.0w-hp-hpux11.00
> 
>   Michael:
> 
>   I also have an hppa2.0w-hp-hpux11.0, but don't have this problem you
>   are referring to.  I think you may have just been using too much memory
>   on your machine.
>   I get:
>   checking the maximum length of command line arguments... 786433
>   with today's head CVS sources.  In general, we can't account
>   for users who are eating up too much memory, and the fact that
>   using the largest command line available is a major performance
>   improvement leads me to the conclusion that this fix is not needed.
>   Feel free to convince me otherwise though.  ;)
> 
>   Sorry,
> 
>   Robert Boehne

-- 
Robert Boehne             Software Engineer
Ricardo Software   Chicago Technical Center
TEL: (630)789-0003 x. 238
FAX: (630)789-0127
email:  address@hidden