NAME

XS::Check - Check XS for some common problems

SYNOPSIS

use FindBin '$Bin';
use XS::Check;
my $check = XS::Check->new ();
$check->check_file ("$Bin/synopsis.xs");

produces output

/usr/home/ben/projects/xs-check/examples/synopsis.xs:3: x not a constant type.
/usr/home/ben/projects/xs-check/examples/synopsis.xs:3: len is not a STRLEN variable (unsigned int ).

(This example is included as synopsis.pl in the distribution.)

VERSION

This documents version 0.14 of XS-Check corresponding to git commit ebf92c9ce84862ecc1565096c1c0b42352e13d78 released on Mon Jul 24 10:02:08 2023 +0900.

DESCRIPTION

This module offers ways to check XS files for some common flaws which we have tripped over.

METHODS

new

my $check = XS::Check->new ();

Make a new XS::Check object. The checks are then run using "check" or "check_file".

Changing where the messages go

The messages from "check" or "check_file" are usually printed using Perl's built-in warn function. If you need to have errors reported some other way, supply a code reference to new with the key reporter as follows:

my $usercheck = XS::Check->new (reporter => sub { print "help!" });

The function you supply is then called back when "check" or "check_file" find something to remark on. The function is called with a hash containing the fields

file

the file name of the file where the error occurred, if using "check_file" or if set with "set_file", otherwise the undefined value (undef),

line

the line number where the error occurred, starting from 1,

message

the message from the module, a text string.

The following example demonstrates a user-defined callback using the message and line fields:

use XS::Check;
my $rchecker = XS::Check->new (reporter => \& reporter);
$rchecker->check ("Perl_croak ('croaking');\n");
sub reporter
{
    my %rstuff = @_;
    print "$rstuff{message} at $rstuff{line}.\n";
}

produces output

Remove the 'Perl_' prefix from Perl_croak at 1.

(This example is included as reporter.pl in the distribution.)

This was added in version 0.07 of the module..

check

$check->check ($xs);

See "SUGGESTIONS" for what this reports.

check_file

$check->check ($xs_file);

Convenience method to read in $xs_file then run "check" on it.

This assumes UTF-8 encoding of $xs_file.

set_file

$check->set_file ($file);

Set the file name for error reporting. Use any false value to clear it. For example:

use XS::Check;
my $check = XS::Check->new ();
my $xs = "Perl_croak (\"frog\")\n";
$check->check ($xs);
$check->set_file ('Yabadabado');
$check->check ($xs);
$check->set_file ('');
$check->check ($xs);

produces output

1: Remove the 'Perl_' prefix from Perl_croak.
Yabadabado:1: Remove the 'Perl_' prefix from Perl_croak.
1: Remove the 'Perl_' prefix from Perl_croak.

(This example is included as set-file.pl in the distribution.)

This was added in version 0.08 of the module.

SUGGESTIONS

This section details the possible suggestions made by the module and the motivations behind them.

Use STRLEN in SvPV

Using an int type for the second argument to SvPV may cause errors on 64-bit Perls, because within the macro the address of the variable is taken, and then it is sent to a Perl function, and if the length doesn't match the length of Perl's STRLEN an error may occur.

Use const char * for return value of SvPV

The pointer returned by SvPV is the actual Perl buffer, not a copy, so unless one actually wants to write into it, it's better to use const char * to make sure one does not overwrite it.

malloc/calloc/realloc/free

The C standard library functions malloc, calloc, realloc, and free should usually be replaced with Newx, Newxz, Renew, and Safefree respectively in Perl XS code, because the C standard library functions may cause "free to wrong pool" errors on multithreaded Windows Perls.

Perl_ prefix

Functions of the form Perl_croak should usually not be used, just croak. The Perl_ prefix functions are the actual functions and croak and other such functions are actually macros, but these macros contain hidden arguments. (The hidden arguments are the pTHX_ and similar things seen in the Perl source code.)

This was added in version 0.04 of the module.

Don't use (void) in arguments

XS functions cannot use the ANSI C (void) to indicate that they do not take any arguments, instead this results in a variable called "void" being created.

This was added in version 0.06 of the module.

Dereferencing av_fetch or hv_fetch

One should not dereference the return value of av_fetch or hv_fetch without checking for NULL (zero pointer) since it is possible to get NULL, for example if an array is created with only a tenth element.

For an extended discussion, see http://blogs.perl.org/users/ben_bullock/2020/02/av-fetch-can-return-null.html.

This check does not actually check that the returned value is checked for non-nullness before being dereferenced, only that there is nothing of the form * av_fetch in your code.

This was added in version 0.09 of the module.

Put whitespace before hash comments

The XS manual suggests putting whitespace before # comments to distinguish them from preprocessor statements.

See https://perldoc.perl.org/perlxs#Inserting-POD,-Comments-and-C-Preprocessor-Directives.

    Comments can be added to XSUBs by placing a # as the first non-whitespace of a line. Care should be taken to avoid making the comment look like a C preprocessor directive, lest it be interpreted as such. The simplest way to prevent this is to put whitespace in front of the #.

This was added in version 0.09 of the module..

Add one to av_len

The av_len function is something of a booby trap in that it returns the length of an AV * minus one, so if it is used as-is, one element of the array will be missed. The module does a simplistic check of seeing whether you have added one to av_len.

This was added in version 0.10 of the module.

Use SvPVbyte or SvPVutf8 rather than SvPV

To work around Perl's strings sometimes being in an ambiguous state, it is better to specify either SvPVbyte or SvPVutf8 and friends rather than plain SvPV. This check is also applied to the similar functions like SvPV_nolen.

The reasons for this are documented in perldoc perlapi for Perl versions from 5.34 onwards.

This was added in version 0.12 of the module.

LIMITATIONS

As of 0.14, the module has the following limitations.

Struct members

The module is not very good at parsing struct members, so XS code like the following doesn't get dealt with properly:

s.txt = SvPV (sv, s.len);
UTF-8 only

"check_file" uses "read_text" in File::Slurper to read the text, which means it only accepts text encoded as UTF-8.

Variable declarations rely on a simplistic hack

The current method of parsing variable declarations uses a very simplistic hack, and it is likely to produce false results if a variable name is used twice for two different things in the same file.

Variables declared within function definitions are not parsed

The following variable length is not dealt with correctly:

static void
sv_to_text_fuzzy (SV * text, STRLEN length)
{
    const unsigned char * stuff;
    /* Copy the string in "text" into "text_fuzzy". */
    stuff = (unsigned char *) SvPV (text, length);

DEPENDENCIES

C::Tokenize

This supplies the regular expressions used to parse C by the module.

"read_text" in File::Slurper

This is used by "check_file".

Text::LineNumber

This is used to get the line numbers.

Carp

COMMAND-LINE TOOL

A command line tool called checkxs is installed with the module. It runs the "check_file" method on each file named on the command line.

checkxs Some.xs

There are two options:

--verbose

Switch on verbose debugging output

--version

Print the version of XS::Check in use.

SEE ALSO

Other CPAN modules

ExtUtils::ParseXS

ExtUtils::ParseXS is Perl's parser for XS which converts XS code into C.

Test::XS::Check

Test::XS::Check is a test module based on this one.

XS::Tutorial

XS::Tutorial is a tutorial about programming in XS.

More information

Perl XS modules and CPAN testers

A collection of more or less obscure bugs found by CPAN testers, the original inspiration for this module.

AUTHOR

Ben Bullock, <bkb@cpan.org>

COPYRIGHT & LICENCE

This package and associated files are copyright (C) 2017-2023 Ben Bullock.

You can use, copy, modify and redistribute this package and associated files under the Perl Artistic Licence or the GNU General Public Licence.