NAME

perlhacktips - Tips for Perl core C code hacking

DESCRIPTION

This document will help you learn the best way to go about hacking on the Perl core C code. It covers common problems, debugging, profiling, and more.

If you haven't read perlhack and perlhacktut yet, you might want to do that first.

COMMON PROBLEMS

Perl source plays by ANSI C89 rules: no C99 (or C++) extensions. In some cases we have to take pre-ANSI requirements into consideration. You don't care about some particular platform having broken Perl? I hear there is still a strong demand for J2EE programmers.

Perl environment problems

  • Not compiling with threading

    Compiling with threading (-Duseithreads) completely rewrites the function prototypes of Perl. You better try your changes with that. Related to this is the difference between "Perl_-less" and "Perl_-ly" APIs, for example:

      Perl_sv_setiv(aTHX_ ...);
      sv_setiv(...);

    The first one explicitly passes in the context, which is needed for e.g. threaded builds. The second one does that implicitly; do not get them mixed. If you are not passing in a aTHX_, you will need to do a dTHX (or a dVAR) as the first thing in the function.

    See "How multiple interpreters and concurrency are supported" in perlguts for further discussion about context.

  • Not compiling with -DDEBUGGING

    The DEBUGGING define exposes more code to the compiler, therefore more ways for things to go wrong. You should try it.

  • Introducing (non-read-only) globals

    Do not introduce any modifiable globals, truly global or file static. They are bad form and complicate multithreading and other forms of concurrency. The right way is to introduce them as new interpreter variables, see intrpvar.h (at the very end for binary compatibility).

    Introducing read-only (const) globals is okay, as long as you verify with e.g. nm libperl.a|egrep -v ' [TURtr] ' (if your nm has BSD-style output) that the data you added really is read-only. (If it is, it shouldn't show up in the output of that command.)

    If you want to have static strings, make them constant:

      static const char etc[] = "...";

    If you want to have arrays of constant strings, note carefully the right combination of consts:

        static const char * const yippee[] =
            {"hi", "ho", "silver"};

    There is a way to completely hide any modifiable globals (they are all moved to heap), the compilation setting -DPERL_GLOBAL_STRUCT_PRIVATE. It is not normally used, but can be used for testing, read more about it in "Background and PERL_IMPLICIT_CONTEXT" in perlguts.

  • Not exporting your new function

    Some platforms (Win32, AIX, VMS, OS/2, to name a few) require any function that is part of the public API (the shared Perl library) to be explicitly marked as exported. See the discussion about embed.pl in perlguts.

  • Exporting your new function

    The new shiny result of either genuine new functionality or your arduous refactoring is now ready and correctly exported. So what could possibly go wrong?

    Maybe simply that your function did not need to be exported in the first place. Perl has a long and not so glorious history of exporting functions that it should not have.

    If the function is used only inside one source code file, make it static. See the discussion about embed.pl in perlguts.

    If the function is used across several files, but intended only for Perl's internal use (and this should be the common case), do not export it to the public API. See the discussion about embed.pl in perlguts.

Portability problems

The following are common causes of compilation and/or execution failures, not common to Perl as such. The C FAQ is good bedtime reading. Please test your changes with as many C compilers and platforms as possible; we will, anyway, and it's nice to save oneself from public embarrassment.

If using gcc, you can add the -std=c89 option which will hopefully catch most of these unportabilities. (However it might also catch incompatibilities in your system's header files.)

Use the Configure -Dgccansipedantic flag to enable the gcc -ansi -pedantic flags which enforce stricter ANSI rules.

If using the gcc -Wall note that not all the possible warnings (like -Wunitialized) are given unless you also compile with -O.

Note that if using gcc, starting from Perl 5.9.5 the Perl core source code files (the ones at the top level of the source code distribution, but not e.g. the extensions under ext/) are automatically compiled with as many as possible of the -std=c89, -ansi, -pedantic, and a selection of -W flags (see cflags.SH).

Also study perlport carefully to avoid any bad assumptions about the operating system, filesystems, and so forth.

You may once in a while try a "make microperl" to see whether we can still compile Perl with just the bare minimum of interfaces. (See README.micro.)

Do not assume an operating system indicates a certain compiler.

  • Casting pointers to integers or casting integers to pointers

        void castaway(U8* p)
        {
          IV i = p;

    or

        void castaway(U8* p)
        {
          IV i = (IV)p;

    Both are bad, and broken, and unportable. Use the PTR2IV() macro that does it right. (Likewise, there are PTR2UV(), PTR2NV(), INT2PTR(), and NUM2PTR().)

  • Casting between data function pointers and data pointers

    Technically speaking casting between function pointers and data pointers is unportable and undefined, but practically speaking it seems to work, but you should use the FPTR2DPTR() and DPTR2FPTR() macros. Sometimes you can also play games with unions.

  • Assuming sizeof(int) == sizeof(long)

    There are platforms where longs are 64 bits, and platforms where ints are 64 bits, and while we are out to shock you, even platforms where shorts are 64 bits. This is all legal according to the C standard. (In other words, "long long" is not a portable way to specify 64 bits, and "long long" is not even guaranteed to be any wider than "long".)

    Instead, use the definitions IV, UV, IVSIZE, I32SIZE, and so forth. Avoid things like I32 because they are not guaranteed to be exactly 32 bits, they are at least 32 bits, nor are they guaranteed to be int or long. If you really explicitly need 64-bit variables, use I64 and U64, but only if guarded by HAS_QUAD.

  • Assuming one can dereference any type of pointer for any type of data

      char *p = ...;
      long pony = *p;    /* BAD */

    Many platforms, quite rightly so, will give you a core dump instead of a pony if the p happens not be correctly aligned.

  • Lvalue casts

      (int)*p = ...;    /* BAD */

    Simply not portable. Get your lvalue to be of the right type, or maybe use temporary variables, or dirty tricks with unions.

  • Assume anything about structs (especially the ones you don't control, like the ones coming from the system headers)

    • That a certain field exists in a struct

    • That no other fields exist besides the ones you know of

    • That a field is of certain signedness, sizeof, or type

    • That the fields are in a certain order

      • While C guarantees the ordering specified in the struct definition, between different platforms the definitions might differ

    • That the sizeof(struct) or the alignments are the same everywhere

      • There might be padding bytes between the fields to align the fields - the bytes can be anything

      • Structs are required to be aligned to the maximum alignment required by the fields - which for native types is for usually equivalent to sizeof() of the field

  • Assuming the character set is ASCIIish

    Perl can compile and run under EBCDIC platforms. See perlebcdic. This is transparent for the most part, but because the character sets differ, you shouldn't use numeric (decimal, octal, nor hex) constants to refer to characters. You can safely say 'A', but not 0x41. You can safely say '\n', but not \012. If a character doesn't have a trivial input form, you can create a #define for it in both utfebcdic.h and utf8.h, so that it resolves to different values depending on the character set being used. (There are three different EBCDIC character sets defined in utfebcdic.h, so it might be best to insert the #define three times in that file.)

    Also, the range 'A' - 'Z' in ASCII is an unbroken sequence of 26 upper case alphabetic characters. That is not true in EBCDIC. Nor for 'a' to 'z'. But '0' - '9' is an unbroken range in both systems. Don't assume anything about other ranges.

    Many of the comments in the existing code ignore the possibility of EBCDIC, and may be wrong therefore, even if the code works. This is actually a tribute to the successful transparent insertion of being able to handle EBCDIC without having to change pre-existing code.

    UTF-8 and UTF-EBCDIC are two different encodings used to represent Unicode code points as sequences of bytes. Macros with the same names (but different definitions) in utf8.h and utfebcdic.h are used to allow the calling code to think that there is only one such encoding. This is almost always referred to as utf8, but it means the EBCDIC version as well. Again, comments in the code may well be wrong even if the code itself is right. For example, the concept of invariant characters differs between ASCII and EBCDIC. On ASCII platforms, only characters that do not have the high-order bit set (i.e. whose ordinals are strict ASCII, 0 - 127) are invariant, and the documentation and comments in the code may assume that, often referring to something like, say, hibit. The situation differs and is not so simple on EBCDIC machines, but as long as the code itself uses the NATIVE_IS_INVARIANT() macro appropriately, it works, even if the comments are wrong.

  • Assuming the character set is just ASCII

    ASCII is a 7 bit encoding, but bytes have 8 bits in them. The 128 extra characters have different meanings depending on the locale. Absent a locale, currently these extra characters are generally considered to be unassigned, and this has presented some problems. This is being changed starting in 5.12 so that these characters will be considered to be Latin-1 (ISO-8859-1).

  • Mixing #define and #ifdef

      #define BURGLE(x) ... \
      #ifdef BURGLE_OLD_STYLE        /* BAD */
      ... do it the old way ... \
      #else
      ... do it the new way ... \
      #endif

    You cannot portably "stack" cpp directives. For example in the above you need two separate BURGLE() #defines, one for each #ifdef branch.

  • Adding non-comment stuff after #endif or #else

      #ifdef SNOSH
      ...
      #else !SNOSH    /* BAD */
      ...
      #endif SNOSH    /* BAD */

    The #endif and #else cannot portably have anything non-comment after them. If you want to document what is going (which is a good idea especially if the branches are long), use (C) comments:

      #ifdef SNOSH
      ...
      #else /* !SNOSH */
      ...
      #endif /* SNOSH */

    The gcc option -Wendif-labels warns about the bad variant (by default on starting from Perl 5.9.4).

  • Having a comma after the last element of an enum list

      enum color {
        CERULEAN,
        CHARTREUSE,
        CINNABAR,     /* BAD */
      };

    is not portable. Leave out the last comma.

    Also note that whether enums are implicitly morphable to ints varies between compilers, you might need to (int).

  • Using //-comments

      // This function bamfoodles the zorklator.   /* BAD */

    That is C99 or C++. Perl is C89. Using the //-comments is silently allowed by many C compilers but cranking up the ANSI C89 strictness (which we like to do) causes the compilation to fail.

  • Mixing declarations and code

      void zorklator()
      {
        int n = 3;
        set_zorkmids(n);    /* BAD */
        int q = 4;

    That is C99 or C++. Some C compilers allow that, but you shouldn't.

    The gcc option -Wdeclaration-after-statements scans for such problems (by default on starting from Perl 5.9.4).

  • Introducing variables inside for()

      for(int i = ...; ...; ...) {    /* BAD */

    That is C99 or C++. While it would indeed be awfully nice to have that also in C89, to limit the scope of the loop variable, alas, we cannot.

  • Mixing signed char pointers with unsigned char pointers

      int foo(char *s) { ... }
      ...
      unsigned char *t = ...; /* Or U8* t = ... */
      foo(t);   /* BAD */

    While this is legal practice, it is certainly dubious, and downright fatal in at least one platform: for example VMS cc considers this a fatal error. One cause for people often making this mistake is that a "naked char" and therefore dereferencing a "naked char pointer" have an undefined signedness: it depends on the compiler and the flags of the compiler and the underlying platform whether the result is signed or unsigned. For this very same reason using a 'char' as an array index is bad.

  • Macros that have string constants and their arguments as substrings of the string constants

      #define FOO(n) printf("number = %d\n", n)    /* BAD */
      FOO(10);

    Pre-ANSI semantics for that was equivalent to

      printf("10umber = %d\10");

    which is probably not what you were expecting. Unfortunately at least one reasonably common and modern C compiler does "real backward compatibility" here, in AIX that is what still happens even though the rest of the AIX compiler is very happily C89.

  • Using printf formats for non-basic C types

       IV i = ...;
       printf("i = %d\n", i);    /* BAD */

    While this might by accident work in some platform (where IV happens to be an int), in general it cannot. IV might be something larger. Even worse the situation is with more specific types (defined by Perl's configuration step in config.h):

       Uid_t who = ...;
       printf("who = %d\n", who);    /* BAD */

    The problem here is that Uid_t might be not only not int-wide but it might also be unsigned, in which case large uids would be printed as negative values.

    There is no simple solution to this because of printf()'s limited intelligence, but for many types the right format is available as with either 'f' or '_f' suffix, for example:

       IVdf /* IV in decimal */
       UVxf /* UV is hexadecimal */
    
       printf("i = %"IVdf"\n", i); /* The IVdf is a string constant. */
    
       Uid_t_f /* Uid_t in decimal */
    
       printf("who = %"Uid_t_f"\n", who);

    Or you can try casting to a "wide enough" type:

       printf("i = %"IVdf"\n", (IV)something_very_small_and_signed);

    Also remember that the %p format really does require a void pointer:

       U8* p = ...;
       printf("p = %p\n", (void*)p);

    The gcc option -Wformat scans for such problems.

  • Blindly using variadic macros

    gcc has had them for a while with its own syntax, and C99 brought them with a standardized syntax. Don't use the former, and use the latter only if the HAS_C99_VARIADIC_MACROS is defined.

  • Blindly passing va_list

    Not all platforms support passing va_list to further varargs (stdarg) functions. The right thing to do is to copy the va_list using the Perl_va_copy() if the NEED_VA_COPY is defined.

  • Using gcc statement expressions

       val = ({...;...;...});    /* BAD */

    While a nice extension, it's not portable. The Perl code does admittedly use them if available to gain some extra speed (essentially as a funky form of inlining), but you shouldn't.

  • Binding together several statements in a macro

    Use the macros STMT_START and STMT_END.

       STMT_START {
          ...
       } STMT_END
  • Testing for operating systems or versions when should be testing for features

      #ifdef __FOONIX__    /* BAD */
      foo = quux();
      #endif

    Unless you know with 100% certainty that quux() is only ever available for the "Foonix" operating system and that is available and correctly working for all past, present, and future versions of "Foonix", the above is very wrong. This is more correct (though still not perfect, because the below is a compile-time check):

      #ifdef HAS_QUUX
      foo = quux();
      #endif

    How does the HAS_QUUX become defined where it needs to be? Well, if Foonix happens to be Unixy enough to be able to run the Configure script, and Configure has been taught about detecting and testing quux(), the HAS_QUUX will be correctly defined. In other platforms, the corresponding configuration step will hopefully do the same.

    In a pinch, if you cannot wait for Configure to be educated, or if you have a good hunch of where quux() might be available, you can temporarily try the following:

      #if (defined(__FOONIX__) || defined(__BARNIX__))
      # define HAS_QUUX
      #endif
    
      ...
    
      #ifdef HAS_QUUX
      foo = quux();
      #endif

    But in any case, try to keep the features and operating systems separate.

Problematic System Interfaces

  • malloc(0), realloc(0), calloc(0, 0) are non-portable. To be portable allocate at least one byte. (In general you should rarely need to work at this low level, but instead use the various malloc wrappers.)

  • snprintf() - the return type is unportable. Use my_snprintf() instead.

Security problems

Last but not least, here are various tips for safer coding.

  • Do not use gets()

    Or we will publicly ridicule you. Seriously.

  • Do not use strcpy() or strcat() or strncpy() or strncat()

    Use my_strlcpy() and my_strlcat() instead: they either use the native implementation, or Perl's own implementation (borrowed from the public domain implementation of INN).

  • Do not use sprintf() or vsprintf()

    If you really want just plain byte strings, use my_snprintf() and my_vsnprintf() instead, which will try to use snprintf() and vsnprintf() if those safer APIs are available. If you want something fancier than a plain byte string, use SVs and Perl_sv_catpvf().

DEBUGGING

You can compile a special debugging version of Perl, which allows you to use the -D option of Perl to tell more about what Perl is doing. But sometimes there is no alternative than to dive in with a debugger, either to see the stack trace of a core dump (very useful in a bug report), or trying to figure out what went wrong before the core dump happened, or how did we end up having wrong or unexpected results.

Poking at Perl

To really poke around with Perl, you'll probably want to build Perl for debugging, like this:

    ./Configure -d -D optimize=-g
    make

-g is a flag to the C compiler to have it produce debugging information which will allow us to step through a running program, and to see in which C function we are at (without the debugging information we might see only the numerical addresses of the functions, which is not very helpful).

Configure will also turn on the DEBUGGING compilation symbol which enables all the internal debugging code in Perl. There are a whole bunch of things you can debug with this: perlrun lists them all, and the best way to find out about them is to play about with them. The most useful options are probably

    l  Context (loop) stack processing
    t  Trace execution
    o  Method and overloading resolution
    c  String/numeric conversions

Some of the functionality of the debugging code can be achieved using XS modules.

    -Dr => use re 'debug'
    -Dx => use O 'Debug'

Using a source-level debugger

If the debugging output of -D doesn't help you, it's time to step through perl's execution with a source-level debugger.

  • We'll use gdb for our examples here; the principles will apply to any debugger (many vendors call their debugger dbx), but check the manual of the one you're using.

To fire up the debugger, type

    gdb ./perl

Or if you have a core dump:

    gdb ./perl core

You'll want to do that in your Perl source tree so the debugger can read the source code. You should see the copyright message, followed by the prompt.

    (gdb)

help will get you into the documentation, but here are the most useful commands:

  • run [args]

    Run the program with the given arguments.

  • break function_name

  • break source.c:xxx

    Tells the debugger that we'll want to pause execution when we reach either the named function (but see "Internal Functions" in perlguts!) or the given line in the named source file.

  • step

    Steps through the program a line at a time.

  • next

    Steps through the program a line at a time, without descending into functions.

  • continue

    Run until the next breakpoint.

  • finish

    Run until the end of the current function, then stop again.

  • 'enter'

    Just pressing Enter will do the most recent operation again - it's a blessing when stepping through miles of source code.

  • print

    Execute the given C code and print its results. WARNING: Perl makes heavy use of macros, and gdb does not necessarily support macros (see later "gdb macro support"). You'll have to substitute them yourself, or to invoke cpp on the source code files (see "The .i Targets") So, for instance, you can't say

        print SvPV_nolen(sv)

    but you have to say

        print Perl_sv_2pv_nolen(sv)

You may find it helpful to have a "macro dictionary", which you can produce by saying cpp -dM perl.c | sort. Even then, cpp won't recursively apply those macros for you.

gdb macro support

Recent versions of gdb have fairly good macro support, but in order to use it you'll need to compile perl with macro definitions included in the debugging information. Using gcc version 3.1, this means configuring with -Doptimize=-g3. Other compilers might use a different switch (if they support debugging macros at all).

Dumping Perl Data Structures

One way to get around this macro hell is to use the dumping functions in dump.c; these work a little like an internal Devel::Peek, but they also cover OPs and other structures that you can't get at from Perl. Let's take an example. We'll use the $a = $b + $c we used before, but give it a bit of context: $b = "6XXXX"; $c = 2.3;. Where's a good place to stop and poke around?

What about pp_add, the function we examined earlier to implement the + operator:

    (gdb) break Perl_pp_add
    Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.

Notice we use Perl_pp_add and not pp_add - see "Internal Functions" in perlguts. With the breakpoint in place, we can run our program:

    (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'

Lots of junk will go past as gdb reads in the relevant source files and libraries, and then:

    Breakpoint 1, Perl_pp_add () at pp_hot.c:309
    309         dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
    (gdb) step
    311           dPOPTOPnnrl_ul;
    (gdb)

We looked at this bit of code before, and we said that dPOPTOPnnrl_ul arranges for two NVs to be placed into left and right - let's slightly expand it:

 #define dPOPTOPnnrl_ul  NV right = POPn; \
                         SV *leftsv = TOPs; \
                         NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0

POPn takes the SV from the top of the stack and obtains its NV either directly (if SvNOK is set) or by calling the sv_2nv function. TOPs takes the next SV from the top of the stack - yes, POPn uses TOPs - but doesn't remove it. We then use SvNV to get the NV from leftsv in the same way as before - yes, POPn uses SvNV.

Since we don't have an NV for $b, we'll have to use sv_2nv to convert it. If we step again, we'll find ourselves there:

    Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
    1669        if (!sv)
    (gdb)

We can now use Perl_sv_dump to investigate the SV:

    SV = PV(0xa057cc0) at 0xa0675d0
    REFCNT = 1
    FLAGS = (POK,pPOK)
    PV = 0xa06a510 "6XXXX"\0
    CUR = 5
    LEN = 6
    $1 = void

We know we're going to get 6 from this, so let's finish the subroutine:

    (gdb) finish
    Run till exit from #0  Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
    0x462669 in Perl_pp_add () at pp_hot.c:311
    311           dPOPTOPnnrl_ul;

We can also dump out this op: the current op is always stored in PL_op, and we can dump it with Perl_op_dump. This'll give us similar output to B::Debug.

    {
    13  TYPE = add  ===> 14
        TARG = 1
        FLAGS = (SCALAR,KIDS)
        {
            TYPE = null  ===> (12)
              (was rv2sv)
            FLAGS = (SCALAR,KIDS)
            {
    11          TYPE = gvsv  ===> 12
                FLAGS = (SCALAR)
                GV = main::b
            }
        }

# finish this later #

SOURCE CODE STATIC ANALYSIS

Various tools exist for analysing C source code statically, as opposed to dynamically, that is, without executing the code. It is possible to detect resource leaks, undefined behaviour, type mismatches, portability problems, code paths that would cause illegal memory accesses, and other similar problems by just parsing the C code and looking at the resulting graph, what does it tell about the execution and data flows. As a matter of fact, this is exactly how C compilers know to give warnings about dubious code.

lint, splint

The good old C code quality inspector, lint, is available in several platforms, but please be aware that there are several different implementations of it by different vendors, which means that the flags are not identical across different platforms.

There is a lint variant called splint (Secure Programming Lint) available from http://www.splint.org/ that should compile on any Unix-like platform.

There are lint and <splint> targets in Makefile, but you may have to diddle with the flags (see above).

Coverity

Coverity (http://www.coverity.com/) is a product similar to lint and as a testbed for their product they periodically check several open source projects, and they give out accounts to open source developers to the defect databases.

cpd (cut-and-paste detector)

The cpd tool detects cut-and-paste coding. If one instance of the cut-and-pasted code changes, all the other spots should probably be changed, too. Therefore such code should probably be turned into a subroutine or a macro.

cpd (http://pmd.sourceforge.net/cpd.html) is part of the pmd project (http://pmd.sourceforge.net/). pmd was originally written for static analysis of Java code, but later the cpd part of it was extended to parse also C and C++.

Download the pmd-bin-X.Y.zip () from the SourceForge site, extract the pmd-X.Y.jar from it, and then run that on source code thusly:

  java -cp pmd-X.Y.jar net.sourceforge.pmd.cpd.CPD --minimum-tokens 100 --files /some/where/src --language c > cpd.txt

You may run into memory limits, in which case you should use the -Xmx option:

  java -Xmx512M ...

gcc warnings

Though much can be written about the inconsistency and coverage problems of gcc warnings (like -Wall not meaning "all the warnings", or some common portability problems not being covered by -Wall, or -ansi and -pedantic both being a poorly defined collection of warnings, and so forth), gcc is still a useful tool in keeping our coding nose clean.

The -Wall is by default on.

The -ansi (and its sidekick, -pedantic) would be nice to be on always, but unfortunately they are not safe on all platforms, they can for example cause fatal conflicts with the system headers (Solaris being a prime example). If Configure -Dgccansipedantic is used, the cflags frontend selects -ansi -pedantic for the platforms where they are known to be safe.

Starting from Perl 5.9.4 the following extra flags are added:

  • -Wendif-labels

  • -Wextra

  • -Wdeclaration-after-statement

The following flags would be nice to have but they would first need their own Augean stablemaster:

  • -Wpointer-arith

  • -Wshadow

  • -Wstrict-prototypes

The -Wtraditional is another example of the annoying tendency of gcc to bundle a lot of warnings under one switch (it would be impossible to deploy in practice because it would complain a lot) but it does contain some warnings that would be beneficial to have available on their own, such as the warning about string constants inside macros containing the macro arguments: this behaved differently pre-ANSI than it does in ANSI, and some C compilers are still in transition, AIX being an example.

Warnings of other C compilers

Other C compilers (yes, there are other C compilers than gcc) often have their "strict ANSI" or "strict ANSI with some portability extensions" modes on, like for example the Sun Workshop has its -Xa mode on (though implicitly), or the DEC (these days, HP...) has its -std1 mode on.

MEMORY DEBUGGERS

NOTE 1: Running under memory debuggers such as Purify, valgrind, or Third Degree greatly slows down the execution: seconds become minutes, minutes become hours. For example as of Perl 5.8.1, the ext/Encode/t/Unicode.t takes extraordinarily long to complete under e.g. Purify, Third Degree, and valgrind. Under valgrind it takes more than six hours, even on a snappy computer. The said test must be doing something that is quite unfriendly for memory debuggers. If you don't feel like waiting, that you can simply kill away the perl process.

NOTE 2: To minimize the number of memory leak false alarms (see "PERL_DESTRUCT_LEVEL" for more information), you have to set the environment variable PERL_DESTRUCT_LEVEL to 2.

For csh-like shells:

    setenv PERL_DESTRUCT_LEVEL 2

For Bourne-type shells:

    PERL_DESTRUCT_LEVEL=2
    export PERL_DESTRUCT_LEVEL

In Unixy environments you can also use the env command:

    env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib ...

NOTE 3: There are known memory leaks when there are compile-time errors within eval or require, seeing S_doeval in the call stack is a good sign of these. Fixing these leaks is non-trivial, unfortunately, but they must be fixed eventually.

NOTE 4: DynaLoader will not clean up after itself completely unless Perl is built with the Configure option -Accflags=-DDL_UNLOAD_ALL_AT_EXIT.

Rational Software's Purify

Purify is a commercial tool that is helpful in identifying memory overruns, wild pointers, memory leaks and other such badness. Perl must be compiled in a specific way for optimal testing with Purify. Purify is available under Windows NT, Solaris, HP-UX, SGI, and Siemens Unix.

Purify on Unix

On Unix, Purify creates a new Perl binary. To get the most benefit out of Purify, you should create the perl to Purify using:

    sh Configure -Accflags=-DPURIFY -Doptimize='-g' \
     -Uusemymalloc -Dusemultiplicity

where these arguments mean:

  • -Accflags=-DPURIFY

    Disables Perl's arena memory allocation functions, as well as forcing use of memory allocation functions derived from the system malloc.

  • -Doptimize='-g'

    Adds debugging information so that you see the exact source statements where the problem occurs. Without this flag, all you will see is the source filename of where the error occurred.

  • -Uusemymalloc

    Disable Perl's malloc so that Purify can more closely monitor allocations and leaks. Using Perl's malloc will make Purify report most leaks in the "potential" leaks category.

  • -Dusemultiplicity

    Enabling the multiplicity option allows perl to clean up thoroughly when the interpreter shuts down, which reduces the number of bogus leak reports from Purify.

Once you've compiled a perl suitable for Purify'ing, then you can just:

    make pureperl

which creates a binary named 'pureperl' that has been Purify'ed. This binary is used in place of the standard 'perl' binary when you want to debug Perl memory problems.

As an example, to show any memory leaks produced during the standard Perl testset you would create and run the Purify'ed perl as:

    make pureperl
    cd t
    ../pureperl -I../lib harness

which would run Perl on test.pl and report any memory problems.

Purify outputs messages in "Viewer" windows by default. If you don't have a windowing environment or if you simply want the Purify output to unobtrusively go to a log file instead of to the interactive window, use these following options to output to the log file "perl.log":

    setenv PURIFYOPTIONS "-chain-length=25 -windows=no \
     -log-file=perl.log -append-logfile=yes"

If you plan to use the "Viewer" windows, then you only need this option:

    setenv PURIFYOPTIONS "-chain-length=25"

In Bourne-type shells:

    PURIFYOPTIONS="..."
    export PURIFYOPTIONS

or if you have the "env" utility:

    env PURIFYOPTIONS="..." ../pureperl ...

Purify on NT

Purify on Windows NT instruments the Perl binary 'perl.exe' on the fly. There are several options in the makefile you should change to get the most use out of Purify:

  • DEFINES

    You should add -DPURIFY to the DEFINES line so the DEFINES line looks something like:

       DEFINES = -DWIN32 -D_CONSOLE -DNO_STRICT $(CRYPT_FLAG) -DPURIFY=1

    to disable Perl's arena memory allocation functions, as well as to force use of memory allocation functions derived from the system malloc.

  • USE_MULTI = define

    Enabling the multiplicity option allows perl to clean up thoroughly when the interpreter shuts down, which reduces the number of bogus leak reports from Purify.

  • #PERL_MALLOC = define

    Disable Perl's malloc so that Purify can more closely monitor allocations and leaks. Using Perl's malloc will make Purify report most leaks in the "potential" leaks category.

  • CFG = Debug

    Adds debugging information so that you see the exact source statements where the problem occurs. Without this flag, all you will see is the source filename of where the error occurred.

As an example, to show any memory leaks produced during the standard Perl testset you would create and run Purify as:

    cd win32
    make
    cd ../t
    purify ../perl -I../lib harness

which would instrument Perl in memory, run Perl on test.pl, then finally report any memory problems.

valgrind

The excellent valgrind tool can be used to find out both memory leaks and illegal memory accesses. As of version 3.3.0, Valgrind only supports Linux on x86, x86-64 and PowerPC. The special "test.valgrind" target can be used to run the tests under valgrind. Found errors and memory leaks are logged in files named testfile.valgrind.

Valgrind also provides a cachegrind tool, invoked on perl as:

    VG_OPTS=--tool=cachegrind make test.valgrind

As system libraries (most notably glibc) are also triggering errors, valgrind allows to suppress such errors using suppression files. The default suppression file that comes with valgrind already catches a lot of them. Some additional suppressions are defined in t/perl.supp.

To get valgrind and for more information see

    http://developer.kde.org/~sewardj/

PROFILING

Depending on your platform there are various ways of profiling Perl.

There are two commonly used techniques of profiling executables: statistical time-sampling and basic-block counting.

The first method takes periodically samples of the CPU program counter, and since the program counter can be correlated with the code generated for functions, we get a statistical view of in which functions the program is spending its time. The caveats are that very small/fast functions have lower probability of showing up in the profile, and that periodically interrupting the program (this is usually done rather frequently, in the scale of milliseconds) imposes an additional overhead that may skew the results. The first problem can be alleviated by running the code for longer (in general this is a good idea for profiling), the second problem is usually kept in guard by the profiling tools themselves.

The second method divides up the generated code into basic blocks. Basic blocks are sections of code that are entered only in the beginning and exited only at the end. For example, a conditional jump starts a basic block. Basic block profiling usually works by instrumenting the code by adding enter basic block #nnnn book-keeping code to the generated code. During the execution of the code the basic block counters are then updated appropriately. The caveat is that the added extra code can skew the results: again, the profiling tools usually try to factor their own effects out of the results.

Gprof Profiling

gprof is a profiling tool available in many Unix platforms, it uses statistical time-sampling.

You can build a profiled version of perl called "perl.gprof" by invoking the make target "perl.gprof" (What is required is that Perl must be compiled using the -pg flag, you may need to re-Configure). Running the profiled version of Perl will create an output file called gmon.out is created which contains the profiling data collected during the execution.

The gprof tool can then display the collected data in various ways. Usually gprof understands the following options:

  • -a

    Suppress statically defined functions from the profile.

  • -b

    Suppress the verbose descriptions in the profile.

  • -e routine

    Exclude the given routine and its descendants from the profile.

  • -f routine

    Display only the given routine and its descendants in the profile.

  • -s

    Generate a summary file called gmon.sum which then may be given to subsequent gprof runs to accumulate data over several runs.

  • -z

    Display routines that have zero usage.

For more detailed explanation of the available commands and output formats, see your own local documentation of gprof.

quick hint:

    $ sh Configure -des -Dusedevel -Doptimize='-pg' && make perl.gprof
    $ ./perl.gprof someprog # creates gmon.out in current directory
    $ gprof ./perl.gprof > out
    $ view out

GCC gcov Profiling

Starting from GCC 3.0 basic block profiling is officially available for the GNU CC.

You can build a profiled version of perl called perl.gcov by invoking the make target "perl.gcov" (what is required that Perl must be compiled using gcc with the flags -fprofile-arcs -ftest-coverage, you may need to re-Configure).

Running the profiled version of Perl will cause profile output to be generated. For each source file an accompanying ".da" file will be created.

To display the results you use the "gcov" utility (which should be installed if you have gcc 3.0 or newer installed). gcov is run on source code files, like this

    gcov sv.c

which will cause sv.c.gcov to be created. The .gcov files contain the source code annotated with relative frequencies of execution indicated by "#" markers.

Useful options of gcov include -b which will summarise the basic block, branch, and function call coverage, and -c which instead of relative frequencies will use the actual counts. For more information on the use of gcov and basic block profiling with gcc, see the latest GNU CC manual, as of GCC 3.0 see

    http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc.html

and its section titled "8. gcov: a Test Coverage Program"

    http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc_8.html#SEC132

quick hint:

    $ sh Configure -des -Dusedevel -Doptimize='-g' \
        -Accflags='-fprofile-arcs -ftest-coverage' \
        -Aldflags='-fprofile-arcs -ftest-coverage' && make perl.gcov
    $ rm -f regexec.c.gcov regexec.gcda
    $ ./perl.gcov
    $ gcov regexec.c
    $ view regexec.c.gcov

MISCELLANEOUS TRICKS

PERL_DESTRUCT_LEVEL

If you want to run any of the tests yourself manually using e.g. valgrind, or the pureperl or perl.third executables, please note that by default perl does not explicitly cleanup all the memory it has allocated (such as global memory arenas) but instead lets the exit() of the whole program "take care" of such allocations, also known as "global destruction of objects".

There is a way to tell perl to do complete cleanup: set the environment variable PERL_DESTRUCT_LEVEL to a non-zero value. The t/TEST wrapper does set this to 2, and this is what you need to do too, if you don't want to see the "global leaks": For example, for "third-degreed" Perl:

        env PERL_DESTRUCT_LEVEL=2 ./perl.third -Ilib t/foo/bar.t

(Note: the mod_perl apache module uses also this environment variable for its own purposes and extended its semantics. Refer to the mod_perl documentation for more information. Also, spawned threads do the equivalent of setting this variable to the value 1.)

If, at the end of a run you get the message N scalars leaked, you can recompile with -DDEBUG_LEAKING_SCALARS, which will cause the addresses of all those leaked SVs to be dumped along with details as to where each SV was originally allocated. This information is also displayed by Devel::Peek. Note that the extra details recorded with each SV increases memory usage, so it shouldn't be used in production environments. It also converts new_SV() from a macro into a real function, so you can use your favourite debugger to discover where those pesky SVs were allocated.

If you see that you're leaking memory at runtime, but neither valgrind nor -DDEBUG_LEAKING_SCALARS will find anything, you're probably leaking SVs that are still reachable and will be properly cleaned up during destruction of the interpreter. In such cases, using the -Dm switch can point you to the source of the leak. If the executable was built with -DDEBUG_LEAKING_SCALARS, -Dm will output SV allocations in addition to memory allocations. Each SV allocation has a distinct serial number that will be written on creation and destruction of the SV. So if you're executing the leaking code in a loop, you need to look for SVs that are created, but never destroyed between each cycle. If such an SV is found, set a conditional breakpoint within new_SV() and make it break only when PL_sv_serial is equal to the serial number of the leaking SV. Then you will catch the interpreter in exactly the state where the leaking SV is allocated, which is sufficient in many cases to find the source of the leak.

As -Dm is using the PerlIO layer for output, it will by itself allocate quite a bunch of SVs, which are hidden to avoid recursion. You can bypass the PerlIO layer if you use the SV logging provided by -DPERL_MEM_LOG instead.

PERL_MEM_LOG

If compiled with -DPERL_MEM_LOG, both memory and SV allocations go through logging functions, which is handy for breakpoint setting.

Unless -DPERL_MEM_LOG_NOIMPL is also compiled, the logging functions read $ENV{PERL_MEM_LOG} to determine whether to log the event, and if so how:

    $ENV{PERL_MEM_LOG} =~ /m/           Log all memory ops
    $ENV{PERL_MEM_LOG} =~ /s/           Log all SV ops
    $ENV{PERL_MEM_LOG} =~ /t/           include timestamp in Log
    $ENV{PERL_MEM_LOG} =~ /^(\d+)/      write to FD given (default is 2)

Memory logging is somewhat similar to -Dm but is independent of -DDEBUGGING, and at a higher level; all uses of Newx(), Renew(), and Safefree() are logged with the caller's source code file and line number (and C function name, if supported by the C compiler). In contrast, -Dm is directly at the point of malloc(). SV logging is similar.

Since the logging doesn't use PerlIO, all SV allocations are logged and no extra SV allocations are introduced by enabling the logging. If compiled with -DDEBUG_LEAKING_SCALARS, the serial number for each SV allocation is also logged.

DDD over gdb

Those debugging perl with the DDD frontend over gdb may find the following useful:

You can extend the data conversion shortcuts menu, so for example you can display an SV's IV value with one click, without doing any typing. To do that simply edit ~/.ddd/init file and add after:

  ! Display shortcuts.
  Ddd*gdbDisplayShortcuts: \
  /t ()   // Convert to Bin\n\
  /d ()   // Convert to Dec\n\
  /x ()   // Convert to Hex\n\
  /o ()   // Convert to Oct(\n\

the following two lines:

  ((XPV*) (())->sv_any )->xpv_pv  // 2pvx\n\
  ((XPVIV*) (())->sv_any )->xiv_iv // 2ivx

so now you can do ivx and pvx lookups or you can plug there the sv_peek "conversion":

  Perl_sv_peek(my_perl, (SV*)()) // sv_peek

(The my_perl is for threaded builds.) Just remember that every line, but the last one, should end with \n\

Alternatively edit the init file interactively via: 3rd mouse button -> New Display -> Edit Menu

Note: you can define up to 20 conversion shortcuts in the gdb section.

Poison

If you see in a debugger a memory area mysteriously full of 0xABABABAB or 0xEFEFEFEF, you may be seeing the effect of the Poison() macros, see perlclib.

Read-only optrees

Under ithreads the optree is read only. If you want to enforce this, to check for write accesses from buggy code, compile with -DPL_OP_SLAB_ALLOC to enable the OP slab allocator and -DPERL_DEBUG_READONLY_OPS to enable code that allocates op memory via mmap, and sets it read-only at run time. Any write access to an op results in a SIGBUS and abort.

This code is intended for development only, and may not be portable even to all Unix variants. Also, it is an 80% solution, in that it isn't able to make all ops read only. Specifically it

  • 1

    Only sets read-only on all slabs of ops at CHECK time, hence ops allocated later via require or eval will be re-write

  • 2

    Turns an entire slab of ops read-write if the refcount of any op in the slab needs to be decreased.

  • 3

    Turns an entire slab of ops read-write if any op from the slab is freed.

It's not possible to turn the slabs to read-only after an action requiring read-write access, as either can happen during op tree building time, so there may still be legitimate write access.

However, as an 80% solution it is still effective, as currently it catches a write access during the generation of Config.pm, which means that we can't yet build perl with this enabled.

The .i Targets

You can expand the macros in a foo.c file by saying

    make foo.i

which will expand the macros using cpp. Don't be scared by the results.

AUTHOR

This document was originally written by Nathan Torkington, and is maintained by the perl5-porters mailing list.