On 19/10/14 02:36, Jeff King wrote:
> On Sat, Oct 18, 2014 at 10:36:12PM +0100, Ramsay Jones wrote:
>
>> I noticed that your 'jk/prune-mtime' branch removes the last caller
>> of the add_object() function; specifically commit 5f78a431a
>> ("reachable: use traverse_commit_list instead of custom walk", 15-10-2014).
>
> Thanks. I usually rely on the compiler to catch any static instances
> that I missed, but of course this one is extern. Did you use an
> automated tool for this
Yes, see below.
> (I know you often catch "X has no declaration;
> should it be static?" with clang,
No, these are caught with sparse.
>but does clang catch this, too?).
>
>> If you need to re-roll those patches, could you please squash this
>> patch into the above commit. (unless you have plans to add some new
>> callers, of course! ;-) ).
>
> Nope, I just didn't notice that I dropped the last caller. I don't think
> we need another re-roll (fingers crossed), but I'd be happy to have this
> on top.
OK, Thanks!
The tool that I use to spot these situations is actually my evolution
of a perl script that Junio sent to the list ages ago! ;-)
I made only minor alterations so that it would work on MinGW 32-bit
(well, actually msysGit, to be more precise), Cygwin 32-bit and 64-bit.
(The 'stop list' of acceptable symbols is _way_ out of date, but the
way I use the script that doesn't matter; still its on my TODO).
I run the script over each branch I build, master then next then pu, and
diff the results (ie ignoring the master base symbols), so I generally
only notice new symbols introduced into pu by new topics. It has been on
my TODO list for a while (OK years!) to go through the output from the
master branch and remove/make static those symbols which shouldn't be
external. (There are currently 53 symbols on 64-bit Linux. The list is
not identical on every platform, so you need to be careful. Also cgit
uses some symbols which would otherwise be static in git! ;-) ).
Ahem, I don't use git to version the script (static-check.pl), so I'm
only reasonable confident that the script I have easily to hand (Linux
Mint 64-bit) is the same on all platforms ... included below.
ATB,
Ramsay Jones
-- >8 --
#!/usr/bin/perl -w
my %defd = ();
my %used = ();
my %def_ok = map { $_ => 1 } qw(
main
alloc_report
have_git_dir
prepare_git_cmd
print_string_list
tm_to_time_t
unsorted_string_list_has_string
xdl_atol
xdl_cha_first
xdl_cha_next
xdl_mmfile_size
xdl_num_out
xdl_recs_cmp
);
for (<*.o>, <*/*.o>, <*/*/*.o>) {
my $obj = $_;
open(I, "-|", qw(nm -gC), $obj) or die;
while () {
unless (/^[0-9a-f ]+([A-Z]) (\S*)$/) {
print STDERR "? $_";
next;
}
next if ($2 =~ /^\.refptr\./);
if (($1 eq "U") || $1 eq "C") {
$used{$2}++;
}
else {
push @{$defd{$obj}}, $2;
}
}
close I;
}
for my $obj (sort keys %defd) {
my $syms = $defd{$obj};
for my $sym (@$syms) {
next if exists $used{$sym} or exists $def_ok{$sym};
print "$obj - $sym\n";
}
}
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html