Karl Williamson [Tue, 31 Dec 2013 15:42:27 +0000 (08:42 -0700)]
regexec.c: Remove redundant test
A string must be in utf8 format if the first code point that came from
it is above 255; therefore it is redundant to test for both.
Martin McGrath [Tue, 31 Dec 2013 16:22:28 +0000 (09:22 -0700)]
PATCH [perl #120901] perlbug.PL - Add to user feedback/docs
Karl Williamson [Tue, 31 Dec 2013 15:28:20 +0000 (08:28 -0700)]
Merge branch into blead that changes \p above-Unicode behavior
This changes the behavior of Unicode property matching for above-Unicode
code points. It changes the format of mktables output tables for binary
properties for efficiency, eliminating a pass while reading them in.
It changes the format somewhat of the output tables when output for
debugging purposes so that they no longer fail some tests.
Karl Williamson [Sun, 29 Dec 2013 04:36:58 +0000 (21:36 -0700)]
Remove no-longer used inversion list function
The function _invlist_invert_prop() is hereby removed. The recent
changes to allow \p{} to match above-Unicode means that no special
handling of properties need be done when inverting.
This function was accessible to XS code that cheated by using #defines
to pretend it was something it wasn't, but it also has been marked
as subject to change since its inception, and never appeared in any
documentation.
Karl Williamson [Thu, 26 Dec 2013 21:01:49 +0000 (14:01 -0700)]
White-space only
This indents various newly-formed blocks (by the previous commit) in
these three files, and reflows lines to fit into 79 columns
Karl Williamson [Wed, 25 Dec 2013 03:11:23 +0000 (20:11 -0700)]
Change format of mktables output binary property tables
mktables now outputs the tables for binary properties as inversion
lists, with a size as the first element. This means simpler handling of
these tables in the core, including removal of an entire pass over them
(it was done just to get the size). These tables are marked as for
internal use by the Perl core only, so their format is changeable at
will.
Karl Williamson [Tue, 24 Dec 2013 03:35:54 +0000 (20:35 -0700)]
Change \p{} matching for above-Unicode code points
http://markmail.org/message/eod7ukhbbh5tnll4 is the beginning of the
thread that led to this commit.
This commit revises the handling of \p{} and \P{} to treat above-Unicode
code points as typical Unicode unassigned ones, and only output a
warning during matching when the answer is arguable under strict Unicode
rules (that is "matched" for \p{}, and "didn't match" for \P{}). The
exception is if the warning category has been made fatal, then it tries
hard to always output the warning. The definition of \p{All} is changed
to be qr/./s, and no warning is issued at all for matching it against
above-Unicode code points.
Karl Williamson [Thu, 19 Dec 2013 05:57:55 +0000 (22:57 -0700)]
regcomp.c: comment typo and rewording
Karl Williamson [Thu, 19 Dec 2013 05:53:46 +0000 (22:53 -0700)]
regcomp.c: Refactor 'if' statement
This refactoring makes it clear that within a (?[]), we don't try to
optimize the [] part. This is for clarity for the future only, as
currently the only changed behavior is if this is being compiled with /l
rules, and (?[]) generates a syntax error under /l.
Karl Williamson [Thu, 19 Dec 2013 05:41:35 +0000 (22:41 -0700)]
Fatalized non-unicode warnings skip regex optimization
This makes sure that fatalized non-unicode warnings actually get output.
For example \p{Line_Break=CR} would normally get optimized into an EXACT
node. But if the user has made non-unicode warnings fatal indicating
they want to be sure not to try to even match such code points, the
optimization is skipped so that the checks are made.
Documentation for this change will be in a future commit.
Karl Williamson [Wed, 27 Nov 2013 19:16:25 +0000 (12:16 -0700)]
mktables: Split off some functionality
This adds a new function that formats a count of code points. Currently
it calls the current function that formats a generic number. A future
commit will change so that the output of the two functions differ. The
reason for this commit is to make that later commit's difference listing
smaller and easier to understand.
Karl Williamson [Wed, 27 Nov 2013 18:39:48 +0000 (11:39 -0700)]
mktables: Add \p{Unicode}
This is a clearer synonym for \p{Any}
Karl Williamson [Wed, 27 Nov 2013 17:59:08 +0000 (10:59 -0700)]
mktables: Separate out defns of \p{Any} and \p{All}
This is in preparation to making them mean different things, in a future
commit
Karl Williamson [Tue, 26 Nov 2013 03:18:31 +0000 (20:18 -0700)]
regcomp.h: Reorder some #defines
There are no logic changes. The previous commit changed the numbers for
some of the bits. This commit re-arranges things so that the #defines
are again in numerical order.
Karl Williamson [Tue, 26 Nov 2013 03:12:33 +0000 (20:12 -0700)]
Re-order some flag bits to avoid potential branches
The ANYOF_INVERT flag is used in every single pattern match of
[bracketed character classes]. With backtracking, this can be a huge
number. All the other flags' uses pale by comparison. I noticed that
by making it the lowest bit, we don't have to use CBOOL, as the only
possibilities are 0 and 1. cBOOL hopefully will be optimized away, but
not always. This commit reorders some of the flag bits to make this one
the lowest, and adds a compile check to make sure it isn't inadvertently
changed.
Karl Williamson [Mon, 30 Dec 2013 22:16:57 +0000 (15:16 -0700)]
Output regex above-Unicode matching in syn strt class
A warning is supposed to be raised under some conditions when matching
an above-Unicode code point against a Unicode property. Prior to this
patch, if the synthetic start class excluded the code point, the warning
would be skipped, even though it was attempted to be matched.
Karl Williamson [Tue, 26 Nov 2013 02:40:12 +0000 (19:40 -0700)]
Convert regnode to a flag for [...]
Prior to this commit, there were 3 types of ANYOF nodes; now there are
two: regular, and one for the synthetic start class (ssc). This commit
converted the third type dealing with warning about matching \p{}
against non-Unicode code points, into using the spare flag bit for ANYOF
nodes.
This allows this bit to apply to ssc ANYOF nodes, whereas previously it
couldn't. There is a bug in which the warning isn't raised if the match
is rejected by the optimizer, because of this inability. This bug will
be fixed in a later commit.
Another option would have been to create a new node-type which was an
ANYOF_SSC_WARN_SUPER node. But this adds extra complications to things;
and we have a spare bit that we might as well use. The comments give
better possibilities for freeing up 2 bits should they be needed.
Karl Williamson [Mon, 30 Dec 2013 22:04:37 +0000 (15:04 -0700)]
regcomp.c: Split #define into two
The syntethic start class regnode (SSC) and a bracketed character class
node share much of the same data structure, including a flags field,
and some of the same flag bits within it. Currently, only
locale-related flags (under /l rules) are the same between the two
during construction of the SSC. But a future commit will introduce
another common flag. This commit creates an extra #define for use where
we want the common flags, while retaining the existing one for use where
we want the locale flags. The new #define is just a copy of the
existing one, to be changed in the future commit.
Karl Williamson [Tue, 26 Nov 2013 02:31:57 +0000 (19:31 -0700)]
mktables: Better comment some variables
Karl Williamson [Fri, 15 Nov 2013 04:12:40 +0000 (21:12 -0700)]
mktables: Calculate debugging information placement
When outputting debugging information under the -annotate option, it's
nice to line up the columns. This commit does a pass through the tables
where the final real data column is variable width so that it can figure
out where to put the debugging info so as almost all of the columns can
be lined up, and not have to be right-shifted because of overlong real
data.
Certain tables prior to this commit had been manually eyeballed and
column information hard-coded in. This is no longer necessary. This
means that one parameter to the write() function is no longer used, and
is removed here.
Karl Williamson [Fri, 15 Nov 2013 02:30:42 +0000 (19:30 -0700)]
mktables: White-space only
Outdent a just-removed block, and better align several other statements
Karl Williamson [Fri, 15 Nov 2013 02:32:44 +0000 (19:32 -0700)]
mktables: Convert to use new function
The previous commit added a new function used in newly added code; this
changes some existing code to use that function
Karl Williamson [Thu, 14 Nov 2013 04:56:31 +0000 (21:56 -0700)]
mktables: Don't change table format with debugging info
The -annotate option to mktables causes it to output extra information
(in the form of comments) to its generated tables to make them human
readable and useful for debugging. Prior to this commit, this caused
the tables' formats to be changed somewhat by causing what normally
ranges to have a line output for each element of the range. This bloats
the tables, and causes UCD.t to fail, as it is looking for a
particular syntax for the tables.
This commit causes the debugging information to be placed separately
but adjacent to the real data. The ranges remain as they would be
without -annotate. This removes the bloat (as the debugging info is
stripped out as the table is read in) and causes UCD.t to pass.
It also allows for the format of the real data to change in a later
commit, and the debugging info can remain relevant.
A future commit will improve the indentation of the comment annotations
Karl Williamson [Tue, 12 Nov 2013 19:09:19 +0000 (12:09 -0700)]
mktables: Improve display of debugging information
Under the -annotate option, mktables outputs the UTF-8 for the printable
characters. This commit adds a non-spacing blank before each such one
that is supposed to combine with its preceding character (marks). This
causes the display of the character to look better.
This necessitated making a local variable more global in scope.
Karl Williamson [Fri, 8 Nov 2013 16:34:54 +0000 (09:34 -0700)]
lib/Unicode/UCD.t: White-space only
Indent a newly formed block
Karl Williamson [Fri, 8 Nov 2013 16:26:51 +0000 (09:26 -0700)]
Add tests for legacy Unicode data files
There are 5 files in lib/unicore/To that may be in direct use by
applications, and which are not used by Perl itself. These have been
changed in an earlier stable release to have comments in them saying,
their use is deprecated, and that Unicode::UCD gives a stable API for
access to the data they contain. However, no warning is given if an
application reads these files, so the deprecation cycle needs to be
quite long. Until we decide to get rid of these files sometime in the
future, we should make sure they exist and are correct. Since they
aren't actually used by Perl, there were no such tests. This commit
adds some tests. It puts them in lib/Unicode/UCD.t, as that required
the least amount of work, as it already has nearly all the
infrastructure required for testing these.
Karl Williamson [Fri, 8 Nov 2013 16:21:11 +0000 (09:21 -0700)]
lib/Unicode/UCD.t: Anchor a couple of regexes
A future commit will need these to be anchored to avoid false positives.
Karl Williamson [Thu, 7 Nov 2013 19:38:31 +0000 (12:38 -0700)]
lib/Unicode/UCD.t: Clarify diagnostic
This diagnostic comes from either of 2 problems, so mention both of
them.
Karl Williamson [Thu, 7 Nov 2013 18:56:09 +0000 (11:56 -0700)]
lib/Unicode/UCD.t: Rename a $variable
This is in preparation for a future commit where the new name makes more
sense.
Karl Williamson [Wed, 6 Nov 2013 17:56:07 +0000 (10:56 -0700)]
Unicode/UCD.t: Add missing 'next' statement
When a test fails, it should do a 'next' to stop processing the current
property.
Karl Williamson [Wed, 6 Nov 2013 05:52:10 +0000 (22:52 -0700)]
mktables: White-space only
Align a few lines to begin on same column which has been outdented so
nothing exceeds 79 columns
Karl Williamson [Wed, 6 Nov 2013 05:33:06 +0000 (22:33 -0700)]
Unicode::UCD: Remove access to some legacy-only properties
Five files are currently being kept around only because they existed
before Unicode::UCD gave access to the properties they define, and some
application programs may rely on their presence, and format. More
compact files have supplanted the use of these files by the Perl core.
Mistakenly, Unicode::UCD gave access to these files via the made-up
property names that they are referred to by in mktables. This was
undocumented. This commit removes this access.
Karl Williamson [Mon, 4 Nov 2013 16:57:29 +0000 (09:57 -0700)]
mktables: Clarify overloaded variable name
The term 'full' is overloaded here in this small section of code. In
some cases it refers to the full case mapping versus the simple case
mapping; in other cases it refers to the full name for a property as
opposed to the abbreviated name. This commit expands each to indicate
which is meant.
Karl Williamson [Sun, 3 Nov 2013 05:22:48 +0000 (23:22 -0600)]
mktables, UCD.t: Fix nits in comments; add comment
Karl Williamson [Tue, 29 Oct 2013 01:49:55 +0000 (19:49 -0600)]
mktables: Don't output trailing tabs in tables
This makes sure that the tabs aren't output unless there is a following
non-null value, saving some disk space
Karl Williamson [Mon, 28 Oct 2013 23:00:25 +0000 (17:00 -0600)]
Unicode/UCD.t: white-space, comments
Wrap to 79 columns; add a comment
Karl Williamson [Mon, 28 Oct 2013 22:43:01 +0000 (16:43 -0600)]
mktables: Stop generating most leading zeros
Leading zeros were generated to conform with Unicode usage, but these
are machine-read files so this just takes up some extra space and extra
parsing cycles at run-time. It's a small matter, but we should design
our files to be the most efficient possible. It is possible to get more
human-readable files by using the -annotate option to mktables.
Certain files whose existence has been published have their formats
unchanged, in case some application is reading them. The files contain
comments that their use is deprecated, but there is no warning generated
if they are opened and read, nor is it really feasible to add such a
warning. At some time in the future, we may feel it's ok to remove
these files, as their contents have been available since v5.16 through a
stable API in Unicode::UCD, but until we remove them, we shouldn't
change their formats.
Not all other leading zeros are removed; just the ones that were
convenient to remove.
Karl Williamson [Sun, 20 Oct 2013 16:57:21 +0000 (10:57 -0600)]
mktables: Further explain how things work in a comment
Karl Williamson [Sun, 20 Oct 2013 16:27:42 +0000 (10:27 -0600)]
mktables: Add an advisory comment to generated files.
Karl Williamson [Sun, 20 Oct 2013 16:20:13 +0000 (10:20 -0600)]
mktables: Regenerate if called with different cmd line args
mktables acts pretty much like its own Makefile. This is because the
rules for regenerating are complicated and too hard to keep in sync in a
Makefile with new versions of Unicode. mktables itself already has
enough intelligence to automatically update the rules when it gets
modified to account for new files from Unicode.
However, prior to this commit, it didn't keep track of the options it
was called with, thus it wouldn't necessarily run when those options
changed to affect the desired outputs.
Karl Williamson [Sun, 20 Oct 2013 16:13:39 +0000 (10:13 -0600)]
mktables: Tighten regex match to real data
The actual file has spaces, so use \s instead of the more dangerous dot.
Also, after processing the line, no need to look to see if it matches
something else.
Karl Williamson [Fri, 18 Oct 2013 02:05:18 +0000 (20:05 -0600)]
mktables: Fixup debugging info
The -annotate parameter generates extra information in the tables
created by mktables which is useful to me in understanding the Unicode
standard and debugging. I doubt that anyone else has ever used it. It
has been broken for some tables for some time. This commit fixes those.
Karl Williamson [Mon, 30 Dec 2013 22:43:12 +0000 (15:43 -0700)]
mktables: Always strip off returned comments in tables
mktables generates (among other things) many perl .pl files which when
executed, return a string containing many lines. Each line may end
with '#' comments. Previously, it didn't always strip off those
comments to the caller, which it assumed uses a 'do' statement to
execute these, and the comments are automatically ignored. However, it
turns out that the 'mkheader' script in Unicode::Normalize doesn't cope
with these comments. This usually doesn't get called except once when
normally these comments aren't generated, but if it does, things don't
just compile. So, just strip off the comments, rather than letting the
'do' handle it.
Karl Williamson [Fri, 18 Oct 2013 02:03:52 +0000 (20:03 -0600)]
mktables: White-space only: wrap to 79 cols
Karl Williamson [Tue, 31 Dec 2013 15:06:17 +0000 (08:06 -0700)]
regcomp.c: Add comments
Karl Williamson [Sun, 29 Dec 2013 04:48:57 +0000 (21:48 -0700)]
regcomp.c: Reinstate use of synthetic start class
This effectively reverts commit
a74bca75951b6a3b0ad03ba07eb31e2ca1227308, although the syntax has
changed. This commit inadvertently caused a synthetic start class (SSC)
to not be generated in many cases where it previously was. The
SSC is generated by the regex optimizer to hopefully speed up finding
where to start matching the target string against the regex pattern. I
don't know if this is a valid data point, but in the 3 months that this
was in blead, there were no complaints of a slowdown that could be
attributed to this.
The commit that caused this was made after discussing it with Yves
Orton. It just seemed (and still seems) wrong that doing what the code
indicates is a logical OR should actually restrict the possibilities.
The change essentially caused the result of OR'ing together the matches
of two nodes, one of which nominally could match a sub-string of zero
length, to also match a sub-string of zero length. Previous to that
commit, and after this new one, the result of doing the OR excludes a
zero-length string, as if it were an AND.
The end result of the change was that the SSC could match a zero-length
string, and thus was discarded as not being useful. Yves and I knew
that the change would not cause bugs; just potentially create more false
positives. And we were right.
I believe that this is further indication that the optimizer could
benefit greatly from an overhaul. It's clear from looking at the code
and commits that other people have been similarly fooled.
Father Chrysostomos [Tue, 31 Dec 2013 05:18:54 +0000 (21:18 -0800)]
toke.c: Remove redundant check
(and turn it into an assertion.)
PL_lex_repl can only be set here if we are parsing a substitution
(OP_SUBST) or transliteration (OP_TRANS). I believe this check
has been unnecessary since perl 5.000, which it was added.
Father Chrysostomos [Mon, 30 Dec 2013 01:11:24 +0000 (17:11 -0800)]
Improve rcatline bad fh warnings
This is not helpful:
append I/O operator() on closed filehandle STDIN at - line 3.
append I/O operator() on closed filehandle STDIN at - line 4.
(Are you trying to call append I/O operator() on dirhandle STDIN?)
That .=<> compiles to a separate op from <> is an implementation
detail, so show ‘readline()’ as we do with <>.
Father Chrysostomos [Mon, 30 Dec 2013 00:43:03 +0000 (16:43 -0800)]
perlhacktips: -Accflags for PERL_DEBUG_READONLY_OPS
It is not clear to all that this has to be specified as
-Accflags=-D... when used on the Configure command line.
Father Chrysostomos [Sun, 29 Dec 2013 01:34:38 +0000 (17:34 -0800)]
perlguts: Make Memory Allocation and PerlIO top-level
These two sections were in the middle of the subroutines section,
but they have little to do with subroutines, whereas the surrounding
sections do pertain to them. I think it was a mistake for them to
have been put under Subroutines to begin with.
Father Chrysostomos [Sun, 29 Dec 2013 01:28:09 +0000 (17:28 -0800)]
Consistent spaces after dots in perlintern.pod
Father Chrysostomos [Sat, 28 Dec 2013 14:55:13 +0000 (06:55 -0800)]
perlapi: Consistent spaces after dots
plus some typo fixes. I probably changed some things in perlintern, too.
Daniel Dragan [Thu, 26 Dec 2013 19:18:16 +0000 (14:18 -0500)]
cut out/inline wrapper calls of sv_*catpvf*
these wrappers use 1 or 2 indirect calls to reach the guts of Perl's
format string engine. Win64 and ARM (and I assume other 4 register regcall
platforms) compilers can't easily tailcall or optimize these due to
sv_vcatpvfn_flags > 4 args and strlen calls. They allocate lots of C
stack and set up new frames and copy args over. So just call the core
format string function directly.
Not all paths to sv_vcatpvfn_flags were optimized since they have no CORE
or CPAN grep usage. sv_vcatpvf has usage, sv_catpvf_mg* have none but were
done anyway because they sound like they would be commonly used.
James E Keenan [Tue, 24 Dec 2013 02:43:16 +0000 (03:43 +0100)]
Add names (descriptions) to t/dumper.t tests lacking them.
TEST() runs over its first argument three times. The description is only
printed on the first of the three tests. Suggestions for more precise
descriptions welcome.
For: RT #120854
Craig A. Berry [Sat, 28 Dec 2013 20:09:46 +0000 (14:09 -0600)]
Fix $Config{src} on VMS.
It may not matter much as it's been broken since
dab31494e8bab and
doesn't seem to have caused any trouble, but to be pedantically
correct the location of the source is not the same as the install
location.
Craig A. Berry [Sat, 28 Dec 2013 17:22:28 +0000 (11:22 -0600)]
Make perl_setup.com infer perl_root from its own location.
The perl_setup.com command procedure initializes the root logical
name, the shareable image logical name pointing to the main Perl
dynamic library, and either command symbols or command table
entries that reference those logical names. It's always by
default hard-coded the prefix chosen at configuration time as the
basis for where Perl should be located.
The original rationale for that was that it gets run right before
installperl and sets things up correctly for installation. But
that optimizes for a single use case that is easily handled by
specifying the correct install location as a parameter and
pessimizes for every other use case, notably relocating an
installed Perl directory tree.
So make it infer the correct root from its own location and
specify the install location only at installation time. Among
other benefits, this allows the install location to be chosen
at installation time by overriding the default prefix like so:
MMK/MACRO=(PREFIX="DSA0:[homedir.MyPerl2.]") install
Nicholas Clark [Fri, 27 Dec 2013 15:22:45 +0000 (16:22 +0100)]
Merge the sfio removal to blead.
Nicholas Clark [Fri, 27 Dec 2013 15:20:08 +0000 (16:20 +0100)]
Add the removal of the sfio support to perldelta.
Given that this code no longer works, it doesn't feel important to list it in
the headlines. It's not clear where the best section for it is.
Nicholas Clark [Tue, 10 Dec 2013 18:32:59 +0000 (19:32 +0100)]
Purge sfio support from configure.com, Configure and related files.
Most of these are actually generated, so the maintenance complexity reduction
is not as impressive as the diffstat suggests.
(Incorporating a fix from Merijn)
Nicholas Clark [Mon, 9 Dec 2013 21:18:22 +0000 (22:18 +0100)]
Purge sfio support, which has been broken for a decade.
The last Perl release that built with -Dusesfio was v5.8.0, and even that
failed many regression tests. Every subsequent release fails to build, and
in the decade that has passed we have had no bug reports about this. So it's
safe to delete all the code. The Configure related code will be purged in a
subsequent commit.
2 references to sfio intentionally remain in fakesdio.h and nostdio.h, as
these appear to be for using its stdio API-compatibility layer.
Nicholas Clark [Mon, 9 Dec 2013 08:15:10 +0000 (09:15 +0100)]
Fix sfio breakage caused by the fix for RT #78494.
Commit
2e0cfa16dea85dd3 (Feb 2011) fixed RT #78494, but inadvertently broke
non-PerlIO configurations. -Uuseperlio was restored in March 2011 with commit
b6ae43b72863e12d, but this change is needed to restore the sfio build.
This is the last fix needed to get blead building with -Dusesfio, after 10
years of breakage. However, many many tests still fail (and one hangs).
Note, specifically this is only tested with sfio97. Perl is making a couple
of cheating assumptions which break with sfio2005, and I've not attempted to
fix Perl's naughtiness to get it to build with current sfio. (Nor has anyone
else in the past 8 years, it would seem. Or even reported a problem.)
Nicholas Clark [Sun, 8 Dec 2013 20:50:53 +0000 (21:50 +0100)]
Fix a typo introduced in Nov 2008 by commit
4ee3916907333148.
One use of dSAVE_ERRNO was mistyped as dSAVED_ERRNO, but as it was in the
sfio-only code it was not spotted.
Nicholas Clark [Sun, 8 Dec 2013 20:11:46 +0000 (21:11 +0100)]
Fix sfio breakage caused by the addition of PERLIO_TERM in Oct 2006.
Commit
6cb8cb214e85c7d6 in Oct 2006 added a new macro PERLIO_TERM, but didn't
provide an appropriate definition for the sfio build. Commit
6b29934bd65555b7
(also Oct 2006) plus this change provides a working PERLIO_TERM for sfio.
Nicholas Clark [Sun, 8 Dec 2013 18:15:25 +0000 (19:15 +0100)]
Fix sfio breakage in PerlIO_findFILE() caused by commit
de009b76d60bdeb8.
One hunk of commit
de009b76d60bdeb8 (April 2005) makes a variable const,
which is then assigned to. This is in sfio-specific code, so was not noticed.
This change was merged to maint-5.8 as part of commit
f1c3982b668c9e30 (Sep
2005).
Nicholas Clark [Sun, 8 Dec 2013 17:47:54 +0000 (18:47 +0100)]
Also build PerlIO_get_layers with -Dusesfio
This extends patch
5fef3b4a7f0c97d3 from April 2003 from fixing the
stdio-only build to also fix the sfio build, which was broken when
PerlIO_get_layers was added (2 commits previously,
39f7a87036eb8d13).
Without this fix, perl would refuse to build with -Dusesfio.
Given that
39f7a87036eb8d13 was merged to maint-5.8, this means that the
-Dusesfio build was broken for v5.8.1 and later on that branch, as well as
for v5.10.0 and all subsequent releases. Given that in the past 10 years,
no-one has reported this, it's safe to assume that no-one considers sfio
support to be important.
Karl Williamson [Fri, 27 Dec 2013 04:43:07 +0000 (21:43 -0700)]
lib/diagnostics.t: Add test
This adds a test so that both =item lines of an adjacent pair get
tested.
Karl Williamson [Fri, 27 Dec 2013 04:50:28 +0000 (21:50 -0700)]
lib/diagnostics.t: Use different test data
This is testing that things work for two adjacent =item lines in
perldiag.pod that share the same description. However a future commit
will change perldiag.pod so that one of the specific adjacent lines
previously used here will be removed. Therefore, choose a different set
of adjacent lines that aren't scheduled to be changed.
This adds a comment in perldiag that the newly chosen adjacent lines are
thus used in diagnostics.t
Father Chrysostomos [Thu, 26 Dec 2013 02:03:16 +0000 (18:03 -0800)]
Increase $constant::VERSION to 1.31
Father Chrysostomos [Thu, 26 Dec 2013 02:02:57 +0000 (18:02 -0800)]
Remove constant.pm-specific behaviour from Internals::SvREADONLY
Some stuff on CPAN is using this undocumented function, so give
constant.pm its own. It is already a core module, depending on
functionality provided by the core solely for its sake; so this
does not really change its relationship to the core.
Father Chrysostomos [Wed, 25 Dec 2013 14:16:31 +0000 (06:16 -0800)]
Revert "[perl #119801] Stop @DB::dbline modifications from crashing"
This reverts commit
c1cec775e9019cc8ae244d4db239a7ea5c0b343e.
See ticket #120864.
David Mitchell [Mon, 2 Dec 2013 15:04:49 +0000 (15:04 +0000)]
[perl #120426] atof() small value rounding errors
For something like 0.153e-305, which is small, but not quite the smallest
number (which is around 2.2e-308), adding extra digits to the fractional part
could cause unnecessary rounding to zero.
From the bug report:
$ echo 0.1530e-305 | perl -e '$v = <STDIN>; print "v=", $v + 0, "\n";'
v=0
$ echo 0.153e-305 | perl -e '$v = <STDIN>; print "v=", $v + 0, "\n";'
v=1.53e-306
This was because 0.1234e-305 is calculated as
1234 / (10^309)
and 10^309 becomes infinity. In these edge cases, repeatedly decrement
the exponent and divide the mantissa by 10 until the exponent becomes in
range; in this case we instead calculate
123 / (10^308)
David Mitchell [Wed, 4 Dec 2013 15:53:51 +0000 (15:53 +0000)]
do cflags on stdout, not stderr
The current UNIX build system does a strange thing to generate the
appropriate command-line to compile a particular src file. It calls
the cflags shell script, which
1) echoes to stdout the command line needed to compile the specified
file (excluding the name of the src file itself), e.g.
cc -c -Dfoo -Wbar ...
2) echoes the same thing to stderr, prefixied with ' CCCMD ='
Make then does
`sh cflags foo.o` foo.c
the cflags output to stdout is captured by the backticks, and is used
by make as the command line to run (with the foo.c appended). This run is
silent. The output to stderr isn't captured, and gets displayed. So the
user sees:
$ make
`sh cflags foo.o` foo.c
CCCMD = cc -c -Dfoo -Wbar ...
...
This is annoying for 2 reasons:
1) you don't get a simple command-line displayed which you could do a
simple cut and paste with (e.g. when you want to recompile a specific
source file, but alter the flags).
2) The make generates output on stderr, even when then there aren't any
errors. So "make 2>errs" can't be used to quickly spot warnings and
errors.
This commit fixes this by making cflags just output the cc command and
flags to stdout, then get Makefile to call it twice, once to echo
the command-line (on stdout), and once to execute it with backticks.
So the make output is now:
$ make
cc -c -Dfoo -Wbar ... foo.c
...
There is some stuff in Makefile.SH related to cross-compiling, which this
commit make have broken. Specifically the CCCMD and CCCMDSRC macros
have been changed in the normal case to remove backticks (and add them to
the make rules instead), but not for the cross compilation route.
The CC* defs in the cross-compilation case have a trailing -I$(CROSS_LIB)
outside of the backticks, which compilates matters.
However, in the subdir Cross/, there appears to be separate (and
divergent) copies of Makefile.SH and cflags, so maybe the files
I edited are no longer used for cross-compilation????
(followup: according to
<
20131204170112.GA2490@iabyn.com>
the cross-compiling stuff I mentioned above has bit-rotted, and I don't
need to worry about it)
David Mitchell [Tue, 3 Dec 2013 17:12:17 +0000 (17:12 +0000)]
cflags/cflags.SH: use '#' comments
These scripts had a mixture of old-style ':' comments and new-style
'#' comments. Since they have both, the old ones can't be needed for
portability reasons, so standardise on the modern form.
The old-style were just too confusing, especially as my syntax highlighter
didn't know about them.
David Mitchell [Tue, 3 Dec 2013 16:17:56 +0000 (16:17 +0000)]
cflags.SH - add commentary
Add comments to the top of cflags.SH to explain what it does.
Also make it output comments at the top of the generated cflags
script explaining that its auto-generated.
Father Chrysostomos [Wed, 25 Dec 2013 13:44:23 +0000 (05:44 -0800)]
doio.c: Stop semop from modifying its argument
Perl_do_semop, which implements the Perl semop function, copies its
second argument to a new struct array, which it passes to the system’s
semop function. It then copies the contents of the struct back into
the argument’s string buffer.
Neither the Darwin nor Linux documentation says that semop modifies
the structs passed to it, and, even if it did happen, perl has never
handle it correctly. It would have to stringify its argument forcibly
(to avoid copying back into a temporary string buffer) and also call
get-magic. And then it would fail with a read-only argument.
Since read-only arguments have always been permitted and the copy
ing-back has never worked correctly, and since this will cause prob-
lems if we upgrade modifications to COW buffers into crashes (the
PERL_DEBUG_READONLY_COW mode I am working in), this commit removes
that code.
Daniel Dragan [Tue, 24 Dec 2013 10:04:55 +0000 (05:04 -0500)]
pp_sys.c remove null checks and locality
pp_enterwrite, EXTEND contains a funccall, dont save gv around it
pp_ioctl, move optype to first place used to reduce liveness
Father Chrysostomos [Tue, 24 Dec 2013 13:44:01 +0000 (05:44 -0800)]
pp_sys.c: More null check removal
Thanks again to Daniel Dragan for pointing out candidates in
<rt-4.0.18-29603-
1387872667-551.120842-15-0@perl.org> (ticket #120842).
Father Chrysostomos [Tue, 24 Dec 2013 06:48:10 +0000 (22:48 -0800)]
Interpret do CORE() as do-file
a96df64385 inadvertently changed it. do-file is how it has alwas been
interpreted, at least as far back as 5.000, as far as I can tell.
Abigail [Tue, 24 Dec 2013 12:51:32 +0000 (13:51 +0100)]
Let's be consistent.
All the sections are of the form '=head2 version - title', except the
one for v5.19.0, which used two dashes.
But no longer.
Father Chrysostomos [Tue, 24 Dec 2013 06:11:16 +0000 (22:11 -0800)]
op.c: newLOOPOP: Remove code supporting do-sub
Father Chrysostomos [Tue, 24 Dec 2013 06:08:26 +0000 (22:08 -0800)]
pp_sys.c: Remove redundant null checks
Nulls only get pushed on to the stack when pp_coreargs uses them to
represent missing optional arguments. Ops that take * prototypes
will have had their arguments passed through rv2gv first, so they
should always be GVs. GvIOn never returns null. When given a GV
argument, it creates a new IO entry and returns that. When given
any other argument it croaks.
Thank you to Daniel Dragan for providing a list of candidate pp
functions in <rt-4.0.18-29609-
1387862001-129.120842-15-0@perl.org>
(ticket #120842).
Karl Williamson [Mon, 23 Dec 2013 18:34:45 +0000 (11:34 -0700)]
pod/perlunicode: White-space only
Commit
1850f57f068545c630b2fb3aa138bbf0fb406e56 placed this verbatim
text heading starting in the wrong column
Father Chrysostomos [Tue, 24 Dec 2013 04:37:18 +0000 (20:37 -0800)]
pp_sys.c: Remove null checks from pp_sockpair
There is actually no way for nulls to reach this code.
Nulls on the stack only happen with pp_coreargs, and only with
ops that have optional arguments, of which socketpair is not one.
GvIOn uses gv_add_by_type, which adds a new IO if there is not
already one, so it will never return null if the GV is not null.
Matthew Horsfall [Sun, 17 Nov 2013 21:25:57 +0000 (16:25 -0500)]
Rearrange dump.c to organize docs. Add some perlapi docs for debug methods.
Father Chrysostomos [Mon, 23 Dec 2013 14:02:23 +0000 (06:02 -0800)]
perldelta: Silence podchecker
Father Chrysostomos [Mon, 23 Dec 2013 13:57:49 +0000 (05:57 -0800)]
pp.c: Remove redundant diag_listed_as
Daniel Dragan [Sun, 22 Dec 2013 05:54:14 +0000 (00:54 -0500)]
test various types of SVs with call_sv
call_sv takes RVs, PVs, CVs, GVs, and an immortal. This isn't well
documented. CVs and immortals can't, or can't easily be tested from
pure perl, so do it from XS. SVt_PVLV with isGV_with_GP is one thing
call_sv takes but is not tested by this commit. Part of [perl #120826] .
Daniel Dragan [Sun, 22 Dec 2013 06:39:09 +0000 (01:39 -0500)]
refactor pp_socket, pp_socketpair, pp_bind
pp_socket: remove unreachable made by commit
9c9f25b8ce
pp_socketpair: increase locality, now gv2/io2 is tested before gv1 is
processed, *v1 vars become non-const to avoid large WS changes in
opening new scope
pp_bind: move op_type's init so it isn't saved by CC across a func call
Daniel Dragan [Mon, 23 Dec 2013 07:11:29 +0000 (02:11 -0500)]
[perl #115736] fix undocumented param from newATTRSUB_flags
flags param was poorly designed and didn't have a formal api. Replace it
with the bool it really is. See #115736 for details.
Dagfinn Ilmari Mannsåker [Thu, 27 Jun 2013 15:48:36 +0000 (16:48 +0100)]
Remove support for "do SUBROUTINE(LIST)"
It's been deprecated (and emitting a warning) since Perl v5.0.0, and
support for it consitutes nearly 3% of the grammar.
Chris 'BinGOs' Williams [Sun, 22 Dec 2013 14:40:52 +0000 (14:40 +0000)]
Update Encode to CPAN version 2.56
[DELTA]
$Revision: 2.55 $ $Date: 2013/09/14 07:51:59 $
! Encode.pm t/Encode.t
Merged RT#91569: [PATCH] decode_utf8 and non-PVs
https://rt.cpan.org/Ticket/Display.html?id=91569
Father Chrysostomos [Tue, 29 Oct 2013 04:59:14 +0000 (21:59 -0700)]
[perl #119801] Stop @DB::dbline modifications from crashing
The cop address for each breakable line was being stored in the IVX
slot of ${"_<$file"}[$line]. This value itself, writable from Perl
space, was being used as the address of the op to be flagged, whenever
a breakpoint was set.
This meant writing to ${"_<$file"}[$line] and assigning a number (like
42) would cause perl to use 42 as an op address, and crash when trying
to flag the op.
Furthermore, since the array holding the lines could outlive the ops,
setting a breakpoint on the op could write to freed memory or to an
unrelated op (even a different type), potentially changing the beha-
viour of unrelated code.
This commit solves those pitfalls by moving breakpoints into a global
breakpoint bitfield. Dbstate ops now have an extra field on the end
holding a sequence number, representing which bit holds the breakpoint
for that op.
Karl Williamson [Sat, 21 Dec 2013 23:48:48 +0000 (16:48 -0700)]
perlio.h: Indent nested #if
I found this helpful in tracking down an issue, being able to
conveniently pair an #endif with its corresponding #if. The rest of the
file could stand this sort of treatment, but not from me now.
Karl Williamson [Sat, 21 Dec 2013 20:04:44 +0000 (13:04 -0700)]
lib/locale.t: Pass AIX failures
Some AIX locales call a no-break space "graphic". Rather than fail AIX
for this, this commit adopts the same approach used for MSWin, which
similarly has most locales containing some errors. And that is to pass
it if any locales at all are correct. Ideally the vendors wouldn't make
broken locales, but this allows us to test for full legality on other
systems.
Karl Williamson [Sat, 21 Dec 2013 20:04:17 +0000 (13:04 -0700)]
lib/locale.t: Typo in comment
Karl Williamson [Sat, 21 Dec 2013 17:49:15 +0000 (10:49 -0700)]
perlunicode: Nits
This mostly makes things you could type in your program display in
constant width. It also adds hyperlinks, and a couple of wording
changes for clarification.
Karl Williamson [Sat, 21 Dec 2013 16:55:32 +0000 (09:55 -0700)]
perlunicode: Note Bidi_Class changed in Unicode 6.3
This property had been stable for many Unicode releases, but it is not
guaranteed to be stable, and in fact changed in 6.3. I left in the
existing list of property values (noting that others have been added)
because they are referred to elsewhere in the pod, and provide a
reasonable example of how multi-valued properties work.
The text used the term "property" to refer to what is properly called
"property values", and used the term "class" to refer to what should
have been called "property". There is another instance of this
confusion of terms in the pod, which will be corrected in a future
commit
Karl Williamson [Fri, 20 Dec 2013 19:10:16 +0000 (12:10 -0700)]
perluniintro: Nits
Craig A. Berry [Sat, 21 Dec 2013 15:59:09 +0000 (09:59 -0600)]
Bump File::Copy to version 2.29.
Craig A. Berry [Sat, 21 Dec 2013 15:48:29 +0000 (09:48 -0600)]
Simplify and clarify VMS specifics in File::Copy.
4c38808d92b95 added some logic to make a "to" path with no
directory component inherit from the current working directory
rather than the directory portion of the "from" path. It also
added a trailing dot to make null file types unambiguous. But
the comments emphasized the latter and made no mentin of the
former, and the implementation was unnecessarily complex.