develooper Front page | perl.perl5.porters | Postings from March 2001

my purify runs with 5.6.1-to-be

Thread Next
From:
Gurusamy Sarathy
Date:
March 14, 2001 17:17
Subject:
my purify runs with 5.6.1-to-be
Message ID:
200103150116.f2F1GCa26620@smtp3.ActiveState.com
FYI, I have just completed my purify runs on maint-5.6/perl@9162
on Solaris 8.  I'm using the following ~/.purify:

    suppress par DLOpEN; load_locale
    suppress umr setcontext; siglongjmp
    suppress umr __big_float_times_power; __decimal_to_binary_integer
    suppress umr __priocntlset; __priocntl
    suppress mlk ...; thr_setspecific
    suppress plk ...; _findbuf; _wrtchk; _fwrite_unlocked; fwrite
    suppress mlk ...; Perl_yyparse

All but the last are needed to suppress problems in the Solaris
libraries.  The last one ignores anything with yyparse() in the
backtrace--I'm ignoring these (as I usually do) as "known" leaks
from compile-time failures.  I've cross-checked that all of these
are in test cases that deliberately introduce syntax errors and
such.

I'm building perl with:

    setenv PURIFYOPTIONS "-chain-length=25"
    setenv PERL_DESTRUCT_LEVEL 4
    sh Configure -Accflags=-DPURIFY -Doptimize=-g -Uusemymalloc \
        -Dusethreads -Duseithreads
    make all test pureperl
    rm -f t/perl
    cd t
    ln -s ../pureperl perl
    ./perl TEST

The only purify error that shows up is in lib/sdbm.t (coredump
accompanied by a purify internal error report in the Duff's device
in ext/SDBM_File/sdbm/pair.c).  Rational has confirmed that it is
a purify bug.

I've also done various runs with:

   * the findleak program I showed
   * an embedding test that does perl_alloc() & co followed by
     eval_pv(argv[1]), followed by perl_destruct() & co in an
     endless loop
   * another embedding test that allocates one interpreter
     and runs eval_pv(argv[1]) repeatedly in an endless loop

None of them show memory growth.

Things are looking pretty good at this point, at least compared to
any previous release of perl. :)


Sarathy
gsar@ActiveState.com

P.S: Note that these tests are all done without changing how the
global destruction happens.  Alan is doing more testing with his
fully refcount-based destruction, which I expect might turn up cases
hidden by the arena cleanup.

Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About