header dependency; Was: [rfc] Re-merged regcache.h patch

header dependency; Was: [rfc] Re-merged regcache.h patch

[Andrew Cagney]

Hello,

When revising the regcache.h patch, I knocked up the attatched (perl
free zone :-) script.  It grubs around the source identifying which .c
files actually use functions/variables from a given header file.

While this script is probably best treated as a novelty - it only just
works for regcache.h - I'm wondering if anyone knows of a GPL package
that does this analysis properly.

	Andrew
#!/bin/sh

parser=/tmp/do-analize-include-parse
refs=/tmp/do-analize-include-refs
files=/tmp/do-analize-include-files
good=/tmp/do-analize-include-good
rm ${parser}
rm ${refs}


# Generate a parser to recognize any reference to a method in "$@"
echo "Creating parse script" 1>&2
cat "$@" | awk '
# external pointer
$0 != n = gensub (/^extern [^\(\*]*\*([_[:alpha:]][_[:alnum:]]*);$/, "\\1", 1) {
    print "/[^[:alnum:]]" n "[[:space:]]*\\[/ { p = 1 } # pointer"
}
$0 != n = gensub (/^extern [^\(]*[^_\([:alnum:]]([_[:alpha:]][_[:alnum:]]*)[[:space:]]\(.*$/, "\\1", 1) {
    print "/[^[:alnum:]]" n "[[:space:]]*\\(/ { p = 1 } # function"
}
' >> ${parser}
cat <<EOF >> ${parser}
p { print FILENAME ":" NR ":" \$0; p = 0; }
EOF
cat ${parser}


# Parse the source files printing out a list of all references
echo "Parsing" 1>&2
find * \
    -name '*-stub.c' -prune \
    -o -name 'nlm' -prune \
    -o -name 'gdbserver' -prune \
    -o -name 'testsuite' -prune \
    -o -name '*\.c' -type f -print | xargs awk -f ${parser} > ${refs}


# Create a summary file of files containing refs
cut -d: -f1 < ${refs} | sort -u > ${files}


# Check to see if the file includes our target
xargs grep -e '#include *"'$1'"' < ${files} | cut -d: -f 1 | sort > ${good}


diff ${good} ${files}
From keiths@cygnus.com Wed Feb 28 07:53:00 2001
From: Keith Seitz <keiths@cygnus.com>
To: Edward Peschko <edwardp@excitehome.net>
Cc: gdb@sourceware.cygnus.com
Subject: Re: using perl in gdbinit
Date: Wed, 28 Feb 2001 07:53:00 -0000
Message-id: <Pine.SOL.3.91.1010228073240.19631D-100000@ryobi.cygnus.com>
References: <20010227160952.A25814@excitehome.net>
X-SW-Source: 2001-02/msg00432.html
Content-length: 1123

On Tue, 27 Feb 2001, Edward Peschko wrote:

> I'd like to use perl in gdbinit to make gdb commands and bind them to a key. 
> So I could say stuff like:
> 
> local($/) = undef;
> my $functions  = `info functions`;

[snip]

I would guess that you should be able to add a perl interpreter to gdb. I 
don't know a whole lot about how perl was written, but if there is a way 
to create your own perl interpreter like this:

   init_perl ()
    {
        g_interp = Perl_New_Interpreter ();
    }

You should be able to do what you want (albeit in a hacky kind of way).

For insight, we added a tcl interpreter and bound tcl's command parser to 
the gdb command "tk". We also have a tcl command, "gdb_cmd", which can be 
used to execute gdb commands in the tcl interpreter.

I could do things like what you want above (although I never have):
(gdb) tk set help [gdb_cmd help]; puts stdout $help
(gdb) tk set functions [gdb_cmd "info func main"]; puts stdout $functions

Of course, you are still leaving yourself at the mercy of gdb's human 
readable output, so you'll need to parse gdb's output to do some things 
well.

Keith

Re: header dependency; Was: [rfc] Re-merged regcache.h patch

[Andrew Cagney]

Andrew Cagney wrote:
> 
> Hello,
> 
> When revising the regcache.h patch, I knocked up the attatched (perl
> free zone :-) script.  It grubs around the source identifying which .c
> files actually use functions/variables from a given header file.
> 
> While this script is probably best treated as a novelty - it only just
> works for regcache.h - I'm wondering if anyone knows of a GPL package
> that does this analysis properly.

To answer my own question, Steven Johnson pointed out that source
navigator can do this.
It creates a database with complete information and you can query
anything - only now I need to read the manual :-)

	Andrew