Nicholas Clark [Mon, 27 Feb 2006 14:19:32 +0000]
There is now the potential for an upgrade from something without an
offset (PVNV or PVMG) to something with one (PVCV, PVFM). Deal with
this correctly.
p4raw-id: //depot/perl@27340
Rafael Garcia-Suarez [Mon, 27 Feb 2006 13:40:44 +0000]
ptr_table_find() is only defined with ithreads
p4raw-id: //depot/perl@27339
H.Merijn Brand [Mon, 27 Feb 2006 13:28:04 +0000]
Workaround for initialization errors on HP's pre-compiler
Do not ask me to explain. This re-definition works on AIX
and HP-UX. Let's await the smokes.
p4raw-id: //depot/perl@27338
Steve Hay [Mon, 27 Feb 2006 12:21:14 +0000]
Fix code-before-declaration error (under VC++) from change #27334
p4raw-link: @27334 on //depot/perl:
b37c2d43c8bccbefe3985273e9661833102148d0
p4raw-id: //depot/perl@27337
Nicholas Clark [Mon, 27 Feb 2006 12:00:12 +0000]
Change 27330 failed to decontaminate SvCUR/GvFLAGS when upgrading an
SvPV acting as a prototype placeholder to a full typeglob.
p4raw-id: //depot/perl@27336
Anno Siegel [Sun, 26 Feb 2006 23:30:01 +0000]
Testing hash key stringification
Message-Id: <
4871E87D-4B59-40A3-A091-A7A3CFA221CF@mailbox.tu-berlin.de>
p4raw-id: //depot/perl@27335
Andy Lester [Fri, 24 Feb 2006 14:54:35 +0000]
Speedups and shrinkages of SvREFCNT_inc
Message-ID: <
20060224205434.GA17867@petdance.com>
p4raw-id: //depot/perl@27334
Dominic Dunlop [Fri, 24 Feb 2006 17:22:54 +0000]
Suppress System V IPC for 64-bit Darwin builds
Message-Id: <
A685C5DB-23CE-430E-95F1-CC7A9480DDF9@mac.com>
p4raw-id: //depot/perl@27333
Zsban Ambrus [Sat, 25 Feb 2006 05:11:03 +0000]
[perl #38633] doc patch: perlfunc "new" section lacks lock and qr
From: Zsban Ambrus (via RT) <perlbug-followup@perl.org>
Message-ID: <rt-3.0.11-38633-130596.7.
22367548359067@perl.org>
p4raw-id: //depot/perl@27332
Rafael Garcia-Suarez [Mon, 27 Feb 2006 07:33:53 +0000]
Fix argument grouping for some macros
p4raw-id: //depot/perl@27331
Nicholas Clark [Sat, 25 Feb 2006 17:16:24 +0000]
GvFLAGS can be overlaid with SvCUR.
p4raw-id: //depot/perl@27330
SADAHIRO Tomoyuki [Sat, 25 Feb 2006 18:16:45 +0000]
Re: [perl #38619] Bug in lc and uc (interaction between UTF-8, substr, and lc/uc)
Message-Id: <
20060225180934.FCC3.BQW10602@nifty.com>
p4raw-id: //depot/perl@27329
Nicholas Clark [Sat, 25 Feb 2006 16:50:15 +0000]
Use xpv_len rather than xpv_cur in GVs to store GvASSIGN_GENERATION.
Assert that GVs do not access SvCUR or SvLEN.
p4raw-id: //depot/perl@27328
Nicholas Clark [Sat, 25 Feb 2006 16:48:46 +0000]
Abstract the specific use of SvCUR in GVs for detecting variables on
both sides of an assignment using GvASSIGN_GENERATION() and
GvASSIGN_GENERATION_set().
p4raw-id: //depot/perl@27327
Nicholas Clark [Sat, 25 Feb 2006 15:38:25 +0000]
xgv_stash can be stored as a union with the NV, reducing the size of
PVGVs by another pointer.
p4raw-id: //depot/perl@27326
Nicholas Clark [Sat, 25 Feb 2006 15:11:28 +0000]
Curiously none of the macros SvIVX, SvUVX or SvNVX are used on
typeglobs.
p4raw-id: //depot/perl@27325
Nicholas Clark [Sat, 25 Feb 2006 14:50:38 +0000]
Amazingly, it seems that none of the other GV specific macros are
accessed unless the GV thinks that it is a GV.
p4raw-id: //depot/perl@27324
Nicholas Clark [Sat, 25 Feb 2006 00:39:30 +0000]
Store GvGP in the SV head union. For all the common lookups [eg GvCV()]
this avoids 1 pointer dereference and the associated risk of a CPU
cache miss. Although this patch looks deceptively small, I fear its
CBV(*) might be rather high.
(* Crack By Volume)
p4raw-id: //depot/perl@27323
Nicholas Clark [Fri, 24 Feb 2006 23:30:50 +0000]
With PERL_DONT_CREATE_GVSV, don't need to create a new GvSV in
Perl_save_scalar.
p4raw-id: //depot/perl@27322
Nicholas Clark [Fri, 24 Feb 2006 22:43:14 +0000]
Given that SvPV on a typeglob is now irrelevant, really we only need
to store the GvGP.
p4raw-id: //depot/perl@27321
Nicholas Clark [Fri, 24 Feb 2006 21:42:33 +0000]
Given that Perl_gp_free() is refcount-dec-and-maybe-free for the glob
pointer, it's inconsistent that it only sets this GV's GvGP to 0 if
this GV happened to have the last reference. Why should this GV care if
it had the last reference? So always set it to 0. It's free. Gone.
p4raw-id: //depot/perl@27320
Nicholas Clark [Fri, 24 Feb 2006 19:32:11 +0000]
Avoid special case SvMAGIC macros just to cope with Perl_sv_unmagic.
p4raw-id: //depot/perl@27319
Nicholas Clark [Fri, 24 Feb 2006 14:19:37 +0000]
Use PVMGs rather than PVGVs to store the names of our variables in the
pad.
p4raw-id: //depot/perl@27313
Nicholas Clark [Fri, 24 Feb 2006 13:59:57 +0000]
Store the stash for our in the magic slot. This will allow us to use
PVMGs in pad names where previously PVGVs were used. In turn, this
gives much greater flexibility for the layout of PVGVs.
p4raw-id: //depot/perl@27312
Steve Peters [Fri, 24 Feb 2006 13:55:23 +0000]
Upgrade to CPAN-1.86
p4raw-id: //depot/perl@27311
Rafael Garcia-Suarez [Fri, 24 Feb 2006 12:52:01 +0000]
Skip unportable test on VMS
p4raw-id: //depot/perl@27309
Nicholas Clark [Fri, 24 Feb 2006 11:52:53 +0000]
Assert that GvSTASH is only called on PVGVs and PVLVs.
p4raw-id: //depot/perl@27307
Nicholas Clark [Fri, 24 Feb 2006 11:52:28 +0000]
Replace usage of GvSTASH for storing the stash of C<our> with
OURSTASH. Set the stash with OURSTASH_SET.
p4raw-id: //depot/perl@27306
Nicholas Clark [Fri, 24 Feb 2006 11:21:06 +0000]
assert that SvMAGIC is never non-NULL for any PVMG used as a pad name.
p4raw-id: //depot/perl@27304
Nicholas Clark [Fri, 24 Feb 2006 10:56:00 +0000]
Redo previous sv.h change.
p4raw-id: //depot/perl@27302
Andy Lester [Thu, 23 Feb 2006 22:53:40 +0000]
Perl_save_aelem can't be const
Message-ID: <
20060224045340.GA6318@petdance.com>
p4raw-id: //depot/perl@27301
Andy Lester [Tue, 21 Feb 2006 00:27:11 +0000]
unused context warnings
Message-ID: <
20060221062711.GA16160@petdance.com>
p4raw-id: //depot/perl@27300
Nicholas Clark [Fri, 24 Feb 2006 10:29:48 +0000]
As SVs can't both be in a pad, and the name of a variable in the pad,
overlay SVs_PADTMP with SVpad_TYPED and SVs_PADMY with SVpad_OUR.
p4raw-id: //depot/perl@27297
Steve Peters [Fri, 24 Feb 2006 02:56:02 +0000]
Revert change #27295, which I thought fixed builds on Win32.
Instead remove &PL_vtbl_glob from globvar.sym.
p4raw-link: @27295 on //depot/perl:
3476b56103cbe13508b1fd6b46ae7b9cb6e0f7ed
p4raw-id: //depot/perl@27296
Steve Peters [Fri, 24 Feb 2006 02:02:14 +0000]
Last instance of PL_vtbl_glob removed. It was missed change #27289.
p4raw-link: @27289 on //depot/perl:
c0c446747ad6c5bde53bc8415ca16ef77f6320f2
p4raw-id: //depot/perl@27295
Nicholas Clark [Thu, 23 Feb 2006 23:45:39 +0000]
Replace direct flags tests & manipulations for SVpad_TYPED and
SVpad_OUR with macros SvPAD_TYPED(), SvPAD_OUR() etc, to abstract
away the flags bits acutally used to store this information.
p4raw-id: //depot/perl@27294
Nicholas Clark [Thu, 23 Feb 2006 22:17:07 +0000]
Document a sixth use for SVf_FAKE
p4raw-id: //depot/perl@27293
Nicholas Clark [Thu, 23 Feb 2006 20:28:34 +0000]
We have sufficient spare bits to store the SV type in body_details,
so add a small sanity check to ensure that we have the array in the
correct order.
p4raw-id: //depot/perl@27292
Nicholas Clark [Thu, 23 Feb 2006 19:40:23 +0000]
Shrink struct body_details and hence sv.o slightly further by using
bitfields.
p4raw-id: //depot/perl@27291
Nicholas Clark [Thu, 23 Feb 2006 19:15:48 +0000]
Using U8 rather than size type shrinks the body_details table.
p4raw-id: //depot/perl@27290
Nicholas Clark [Thu, 23 Feb 2006 18:00:19 +0000]
Remove set magic from typeglobs. Remove typeglob magic entirely.
Typeglobs now never access the SvPVX, SvIVX or SvNVX when holding a
valid GvGP().
p4raw-id: //depot/perl@27289
Nicholas Clark [Thu, 23 Feb 2006 17:45:55 +0000]
In XS_attributes__guess_stash, attempting to call Gv* on a PVMG is
bad and wrong.
p4raw-id: //depot/perl@27288
Steve Peters [Thu, 23 Feb 2006 17:42:09 +0000]
Add test for RT #2166: foreach spuriously autovivifies
p4raw-id: //depot/perl@27287
Steve Peters [Thu, 23 Feb 2006 16:24:19 +0000]
Converted t/op/pos.t to use test.pl and added TODO test for
RT #1716 - search position reset after 'local' save/restore
p4raw-id: //depot/perl@27286
Jan Dubois [Tue, 21 Feb 2006 12:36:31 +0000]
Implement ${^WIN32_SLOPPY_STAT}
From: "Jan Dubois" <jand@ActiveState.com>
Message-ID: <
019601c63726$7fcca200$6062a8c0@candy>
p4raw-id: //depot/perl@27283
Nicholas Clark [Thu, 23 Feb 2006 13:42:29 +0000]
Recalculate dstr in Perl_sv_setsv_flags, as dstr may have been upgraded.
p4raw-id: //depot/perl@27282
Nicholas Clark [Thu, 23 Feb 2006 13:07:54 +0000]
Remove un-needed case in Perl_sv_setsv_flags (sv_upgrade will cover
this one for us).
p4raw-id: //depot/perl@27279
Nicholas Clark [Thu, 23 Feb 2006 11:11:12 +0000]
Remove get magic from typeglobs. This means that PVGVs holding
typeglobs never need to use SvPVX. This comes at price - typeglobs
were using magic get for their stringificiation, and to pass SvOK(),
so need to make typeglobs SvOK by default (by sucking SVp_SCREAM into
SVf_OK - it's the only flag left), tweak SvSCREAM() to also check
SVp_POK, and teach sv_2[inpu]v how to convert globs.
However, it should free up SvPVX for the next part of the plan to
pointer indirections, and therefore CPU cache pressure.
p4raw-id: //depot/perl@27278
Nicholas Clark [Thu, 23 Feb 2006 00:29:04 +0000]
Fix the copy sizes for PVFM and PVCV back to sanity.
p4raw-id: //depot/perl@27276
Nicholas Clark [Wed, 22 Feb 2006 23:43:10 +0000]
Fix typos and a missing bracket.
p4raw-id: //depot/perl@27274
Nicholas Clark [Wed, 22 Feb 2006 22:30:19 +0000]
Avoid C<study>ing any strings that might change underneath us, such
as tied scalars and scalars with overloaded stringification.
p4raw-id: //depot/perl@27273
Rafael Garcia-Suarez [Wed, 22 Feb 2006 22:09:41 +0000]
Mark some new study tests as TODO
p4raw-id: //depot/perl@27272
Rafael Garcia-Suarez [Wed, 22 Feb 2006 21:04:33 +0000]
Add a test for study() on tied scalars, by Andy Lester after
Rick Delaney
p4raw-id: //depot/perl@27271
Jarkko Hietaniemi [Wed, 22 Feb 2006 20:16:00 +0000]
Turn on match string copying when /e flag is set on a substitution.
Subject: [PATCH] dodge a valgrind error (for maint or blead)
Message-ID: <
43FCB896.7060106@gmail.com>
p4raw-id: //depot/perl@27270
Nicholas Clark [Wed, 22 Feb 2006 18:21:30 +0000]
sv_dump should report the PV for PVGVs, as it can get set.
(Typeglobs stringify via a call to GET magic, which uses the PV slot.)
p4raw-id: //depot/perl@27269
Nicholas Clark [Wed, 22 Feb 2006 16:50:25 +0000]
Test warnings for converting globs to other forms.
p4raw-id: //depot/perl@27268
Nicholas Clark [Wed, 22 Feb 2006 00:23:49 +0000]
Change 24643 made the mistake of assuming that CvCONST can only be true
on XSUBs. Somehow it can also end up on perl subs. Bug spotted by and
test case from Marcus Holland-Moritz.
p4raw-id: //depot/perl@27267
Nicholas Clark [Tue, 21 Feb 2006 20:09:25 +0000]
Can merge the two arms of Perl_magic_getglob to save space.
p4raw-id: //depot/perl@27265
Nicholas Clark [Tue, 21 Feb 2006 19:04:39 +0000]
Reorder the SV flags so that SVp_?OK and SVf_?OK occupy 8 contiguous
bits (0x0000XX00). This makes perl 3K smaller with -Os on x86 FreeBSD,
and might also help on other architectures (eg with 8 contiguous bits,
the SvOK() test on ARM won't need an intermediate constant).
p4raw-id: //depot/perl@27264
Steve Peters [Tue, 21 Feb 2006 18:44:21 +0000]
Additional hints needed to build threaded Perls on NetBSD.
p4raw-id: //depot/perl@27263
Nicholas Clark [Tue, 21 Feb 2006 18:00:57 +0000]
PL_body_arenas should be set to NULL when the interpreter is cleared.
p4raw-id: //depot/perl@27262
Nicholas Clark [Tue, 21 Feb 2006 16:28:02 +0000]
Teach B about CVf_ISXSUB
p4raw-id: //depot/perl@27261
Nicholas Clark [Tue, 21 Feb 2006 16:12:37 +0000]
Goodbye PERL_XSUB_OLDSTYLE.
p4raw-id: //depot/perl@27260
Nicholas Clark [Tue, 21 Feb 2006 15:40:04 +0000]
Re-order the definitions of the private bits in SvFLAGS by value, so
that we can see where we're already double-booked.
p4raw-id: //depot/perl@27259
Nicholas Clark [Tue, 21 Feb 2006 15:23:21 +0000]
Document the many uses of SVf_FAKE.
p4raw-id: //depot/perl@27258
Steve Peters [Tue, 21 Feb 2006 15:18:58 +0000]
Add tests for the previously untested Hash::Util::all_keys().
p4raw-id: //depot/perl@27257
Steve Peters [Tue, 21 Feb 2006 00:43:43 +0000]
$ExtUtils::MM_Unix::VERSION needs to stay numeric to avoid test
warnings.
p4raw-id: //depot/perl@27256
Joshua ben Jore [Sun, 19 Feb 2006 02:58:10 +0000]
Patches: B, CGI, ExtUtils::MM_Unix
From: "Joshua ben Jore" <twists@gmail.com>
Message-ID: <
dc5c751d0602190058t78d915fv78c6318370307b63@mail.gmail.com>
p4raw-id: //depot/perl@27255
Joshua ben Jore [Mon, 20 Feb 2006 09:50:15 +0000]
Re: Patches: B, CGI, ExtUtils::MM_Unix
From: "Joshua ben Jore" <twists@gmail.com>
Message-ID: <
dc5c751d0602200750j21447031m86ea670b04ac27ed@mail.gmail.com>
Includes changes to increment test count in ext/B/t/concise-xs.t and
uses "no warnings 'once';" in ext/B/t/b.t rather than kludging to
avoid the warning.
p4raw-id: //depot/perl@27254
Yitzchak Scott-Thoennes [Mon, 20 Feb 2006 11:09:33 +0000]
op/magic failure on cygwin after 1.5.19-4
Message-ID: <
20060220190933.GA1316@efn.org>
p4raw-id: //depot/perl@27252
Nicholas Clark [Mon, 20 Feb 2006 20:32:09 +0000]
Steal code from maint, and use PERL_ARENA_ROOTS_SIZE to size the arena
arrays, rather than SVt_LAST, so that SVt_LAST can truthfully remain
the number of genuine SV types, unclouded by implementation details.
p4raw-id: //depot/perl@27251
Dominic Dunlop [Mon, 20 Feb 2006 13:46:00 +0000]
Make SDBM_File work with -Duse64bitall on Darwin (Mac OS X)
Message-Id: <
B699DDDE-EC4E-495D-AFF5-3D7399332A24@mac.com>
p4raw-id: //depot/perl@27250
Nicholas Clark [Mon, 20 Feb 2006 19:00:22 +0000]
PVCVs don't need XNVs either.
(And actually remove xcv_depth)
(And fix the copy lengths in bodies_by_type)
p4raw-id: //depot/perl@27249
Dominic Dunlop [Mon, 20 Feb 2006 14:49:00 +0000]
Trouble with $ENV{CDPATH} after change #27236
Message-Id: <
6393FA5A-6B84-46E9-A557-DED3BB0AD7EE@mac.com>
p4raw-id: //depot/perl@27248
Nicholas Clark [Mon, 20 Feb 2006 17:48:21 +0000]
PVFMs don't need CvDEPTH, and PVCVs don't use SvIVX, so moving
xcv_depth into the IV union saves 4(ish) bytes per CV and format.
"ish" because it was a long, but has been changed to I32 (along with
the corresponding field in struct block_sub) so as not to enlarge the
IV union on platforms where sizeof(long) > sizeof(IV), or struct
block_sub where sizeof(long) > sizeof(I32)
p4raw-id: //depot/perl@27247
Nicholas Clark [Mon, 20 Feb 2006 14:01:59 +0000]
Fix sv_dump to dump formats without the (non-existent) NVX.
p4raw-id: //depot/perl@27245
Nicholas Clark [Mon, 20 Feb 2006 13:42:47 +0000]
xcv_root and xcv_xsub can also be merged into a union, providing a new
flag is added to denote whether the PVCV is perl or XSUB.
p4raw-id: //depot/perl@27244
Nicholas Clark [Mon, 20 Feb 2006 11:54:38 +0000]
xcv_start and xcv_xsubany can be merged into a union, as they are never
both needed.
p4raw-id: //depot/perl@27243
Nicholas Clark [Mon, 20 Feb 2006 11:54:03 +0000]
Need to clear CvXSUBANY() too to turn a constant sub into a prototype.
p4raw-id: //depot/perl@27242
Nicholas Clark [Mon, 20 Feb 2006 10:40:59 +0000]
Add a new CvISXSUB() macro, for abstracting the test as to whether a
PVCV is perl or XS.
p4raw-id: //depot/perl@27241
Nicholas Clark [Mon, 20 Feb 2006 10:10:11 +0000]
Abolish BROKEN_UNION_INIT in B::C, as it works around problems in
pre-ANSI C compilers, but pays in code duplication.
p4raw-id: //depot/perl@27240
John E. Malmberg [Mon, 20 Feb 2006 03:43:00 +0000]
patch@27236 vms glob/readdir/chdir EFS/long filename support
Message-ID: <
43F92CE6.5040704@qsl.net>
p4raw-id: //depot/perl@27239
Rafael Garcia-Suarez [Mon, 20 Feb 2006 09:26:15 +0000]
More NullXXX macro removal from Andy Lester
p4raw-id: //depot/perl@27238
Rafael Garcia-Suarez [Mon, 20 Feb 2006 08:54:22 +0000]
Yitzchak points out that the perldiag entry for "Integer overflow in
division" is no longer useful.
p4raw-id: //depot/perl@27237
Nicholas Clark [Mon, 20 Feb 2006 00:36:34 +0000]
utftaint.t won't be able to run tests with -T if you have
Insecure directory in $ENV{PATH}
so skip them. (Probably this means you have . in your PATH)
p4raw-id: //depot/perl@27236
Nicholas Clark [Mon, 20 Feb 2006 00:21:09 +0000]
Remove the last (U16) cast for CV depths, missed by change 17835.
p4raw-id: //depot/perl@27235
Nicholas Clark [Mon, 20 Feb 2006 00:02:53 +0000]
Re-order CV flags to bring the 4 CVf_BUILTIN_ATTRS into adjacent bits,
and make other flag bits that are paired in the code adjacent.
Will produce tighter code on ARM; might help on other platforms too.
p4raw-id: //depot/perl@27234
Nicholas Clark [Sun, 19 Feb 2006 23:59:52 +0000]
Ministry of Truth removes typo in description of Change 26166.
Change 26166 was correct from the start.
Nothing to see. Move along.
We thank Big Brother for raising the chocolate ration to 20g per week.
p4raw-id: //depot/perl@27233
Nicholas Clark [Sun, 19 Feb 2006 22:35:24 +0000]
Shave sizeof(NV) bytes from formats, by using the same offset
manoeuvre as PVs, PVIVs, PVAVs and PVHVs.
p4raw-id: //depot/perl@27231
Nicholas Clark [Sun, 19 Feb 2006 20:30:24 +0000]
AVs and HVs don't have IVXs or NVXs, so assert this too.
p4raw-id: //depot/perl@27230
Nicholas Clark [Sun, 19 Feb 2006 20:10:34 +0000]
Under -DDEBUGGING, assert that SvIVX, SvUVX and SvNVX aren't being
used on scalars that don't have the memory allocated. Correct
SvSTASH and SvMAGIC to only evaluate the sv argument once.
p4raw-id: //depot/perl@27229
Nicholas Clark [Sun, 19 Feb 2006 18:57:35 +0000]
To make arithmetic on tainted dualvars work properly requires that
sv_2nv uses SvIVX in preference to SvPVX, if SVp_IOK is true.
p4raw-id: //depot/perl@27228
Nicholas Clark [Sun, 19 Feb 2006 12:14:15 +0000]
SvROK(sv) will never be true when SvIOKp() or SvNOKp() is true, so the
code inside the if() test in sv_2pv_flags is dead and can be removed.
p4raw-id: //depot/perl@27227
Rafael Garcia-Suarez [Sat, 18 Feb 2006 20:13:51 +0000]
As Anno Siegel points out, mtime/ctime test should be skipped
on OS X on HFS+.
p4raw-id: //depot/perl@27223
Nicholas Clark [Sat, 18 Feb 2006 19:01:48 +0000]
save_re_context() and even errsv_save = newSVsv(ERRSV); can turn
PL_tainted back on, so defer turning if off as late as possible.
This gets lib/locale.t working once more.
p4raw-id: //depot/perl@27222
Nicholas Clark [Sat, 18 Feb 2006 18:23:43 +0000]
Turn of tainting locally inside swash_init(), as the internal
implementation of the regexp engine has nothing to do with user data,
so when and where it happens to call require is agnostic of the current
expression's taintedness.
p4raw-id: //depot/perl@27221
Nicholas Clark [Sat, 18 Feb 2006 18:22:13 +0000]
Assume that if runperl is called under tainting, that the caller really
really wanted to run perl, so brute force untaint everything.
p4raw-id: //depot/perl@27220
Nicholas Clark [Sat, 18 Feb 2006 17:38:38 +0000]
Convert utftaint.t to test.pl (from Test.pm) and provide it with
tainted() from taint.t to remove the dependency on Scalar::Util.
(So it will now work with miniperl)
p4raw-id: //depot/perl@27219
Steve Peters [Sat, 18 Feb 2006 12:58:49 +0000]
Upgrade to Digest-SHA-5.34.
p4raw-id: //depot/perl@27218
Jim Cromie [Mon, 13 Feb 2006 14:12:41 +0000]
arena-rework : consolidated patch
Message-ID: <
43F0F649.9040205@gmail.com>
Tweaked somewhat to split the arena boolean from the arena_size,
and with the PTE still doubling-up with one of the SV types in the
array.
p4raw-id: //depot/perl@27215