Mercurial > pidgin.yaz
view plugins/perl/perl-common.c @ 6561:33ceba0dfd9b
[gaim-migrate @ 7083]
Fix for bug #792641 - Windows hooks tie up
committer: Tailor Script <tailor@pidgin.im>
author | Herman Bloggs <hermanator12002@yahoo.com> |
---|---|
date | Thu, 21 Aug 2003 18:29:48 +0000 |
parents | 7c42b8ca3222 |
children | f6c2a7b5afa7 |
line wrap: on
line source
#include "debug.h" #include "perl-common.h" extern PerlInterpreter *my_perl; static GHashTable *object_stashes = NULL; static int magic_free_object(pTHX_ SV *sv, MAGIC *mg) { sv_setiv(sv, 0); return 0; } static MGVTBL vtbl_free_object = { NULL, NULL, NULL, NULL, magic_free_object }; static SV * create_sv_ptr(void *object) { SV *sv; sv = newSViv((IV)object); sv_magic(sv, NULL, '~', NULL, 0); SvMAGIC(sv)->mg_private = 0x1551; /* HF */ SvMAGIC(sv)->mg_virtual = &vtbl_free_object; return sv; } SV * gaim_perl_bless_object(void *object, const char *stash_name) { HV *stash; HV *hv; if (object_stashes == NULL) { object_stashes = g_hash_table_new(g_direct_hash, g_direct_equal); } stash = gv_stashpv(stash_name, 1); hv = newHV(); hv_store(hv, "_gaim", 5, create_sv_ptr(object), 0); return sv_bless(newRV_noinc((SV *)hv), stash); // return sv_bless(create_sv_ptr(object), gv_stashpv(stash, 1)); // return create_sv_ptr(object); // dXSARGS; // return sv_setref_pv(ST(0), "Gaim::Account", create_sv_ptr(object)); } gboolean gaim_perl_is_ref_object(SV *o) { SV **sv; HV *hv; hv = hvref(o); if (hv != NULL) { sv = hv_fetch(hv, "_gaim", 5, 0); if (sv != NULL) return TRUE; } return FALSE; } void * gaim_perl_ref_object(SV *o) { SV **sv; HV *hv; void *p; hv = hvref(o); if (hv == NULL) return NULL; sv = hv_fetch(hv, "_gaim", 5, 0); if (sv == NULL) croak("variable is damaged"); p = GINT_TO_POINTER(SvIV(*sv)); return p; } /* 2003/02/06: execute_perl modified by Mark Doliner <mark@kingant.net> Pass parameters by pushing them onto the stack rather than passing an array of strings. This way, perl scripts can modify the parameters and we can get the changed values and then shoot ourselves. I mean, uh, use them. 2001/06/14: execute_perl replaced by Martin Persson <mep@passagen.se> previous use of perl_eval leaked memory, replaced with a version that uses perl_call instead 30/11/2002: execute_perl modified by Eric Timme <timothy@voidnet.com> args changed to char** so that we can have preparsed arguments again, and many headaches ensued! This essentially means we replaced one hacked method with a messier hacked method out of perceived necessity. Formerly execute_perl required a single char_ptr, and it would insert it into an array of character pointers and NULL terminate the new array. Now we have to pass in pre-terminated character pointer arrays to accomodate functions that want to pass in multiple arguments. Previously arguments were preparsed because an argument list was constructed in the form 'arg one','arg two' and was executed via a call like &funcname(arglist) (see .59.x), so the arglist was magically pre-parsed because of the method. With Martin Persson's change to perl_call we now need to use a null terminated list of character pointers for arguments if we wish them to be parsed. Lacking a better way to allow for both single arguments and many I created a NULL terminated array in every function that called execute_perl and passed that list into the function. In the former version a single character pointer was passed in, and was placed into an array of character pointers with two elements, with a NULL element tacked onto the back, but this method no longer seemed prudent. Enhancements in the future might be to get rid of pre-declaring the array sizes? I am not comfortable enough with this subject to attempt it myself and hope it to stand the test of time. */ int execute_perl(const char *function, int argc, char **args) { int count = 0, i, ret_value = 1; SV *sv_args[argc]; STRLEN na; /* * Set up the perl environment, push arguments onto the * perl stack, then call the given function */ dSP; ENTER; SAVETMPS; PUSHMARK(sp); for (i = 0; i < argc; i++) { if (args[i]) { sv_args[i] = sv_2mortal(newSVpv(args[i], 0)); XPUSHs(sv_args[i]); } } PUTBACK; count = call_pv(function, G_EVAL | G_SCALAR); SPAGAIN; /* * Check for "die," make sure we have 1 argument, and set our * return value. */ if (SvTRUE(ERRSV)) { gaim_debug(GAIM_DEBUG_ERROR, "perl", "Perl function %s exited abnormally: %s\n", function, SvPV(ERRSV, na)); POPs; } else if (count != 1) { /* * This should NEVER happen. G_SCALAR ensures that we WILL * have 1 parameter. */ gaim_debug(GAIM_DEBUG_ERROR, "perl", "Perl error from %s: expected 1 return value, " "but got %d\n", function, count); } else ret_value = POPi; /* Check for changed arguments */ for (i = 0; i < argc; i++) { if (args[i] && strcmp(args[i], SvPVX(sv_args[i]))) { /* * Shizzel. So the perl script changed one of the parameters, * and we want this change to affect the original parameters. * args[i] is just a tempory little list of pointers. We don't * want to free args[i] here because the new parameter doesn't * overwrite the data that args[i] points to. That is done by * the function that called execute_perl. I'm not explaining this * very well. See, it's aggregate... Oh, but if 2 perl scripts * both modify the data, _that's_ a memleak. This is really kind * of hackish. I should fix it. Look how long this comment is. * Holy crap. */ args[i] = g_strdup(SvPV(sv_args[i], na)); } } PUTBACK; FREETMPS; LEAVE; return ret_value; }