summaryrefslogtreecommitdiff
path: root/embed.fnc
Commit message (Collapse)AuthorAgeFilesLines
* don't check format args on taint_properDavid Mitchell2013-12-011-1/+1
| | | | | | | | My recent commit 5d37acd6b65eb enabled (among other things) format-arg checking of taint_proper(). This was not a good idea since taint_proper() adds extra args before it actually calls a printf-style function. This was masked since on some gcc systems, a NULLOK format arg disables this check.
* silence -Wformat-nonliteral compiler warningsDavid Mitchell2013-11-281-2/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Due to the security risks associated with user-supplied formats being passed to C-level printf() style functions (eg %n), gcc has a -Wformat-nonliteral warning that complains whenever such a function is passed a non-literal format string. This commit silences all such warnings in core and ext/. The main changes are 1) the 'f' (format) flag in embed.fnc is now handled slightly more cleverly. Rather than just applying to functions whose last arg is '...' (and where the format arg is assumed to be the previous arg), it can now handle non-'...' functions: arg checking is disabled, but format checking is sill done: it works by assuming that an arg called 'fmt', 'pat' or 'f' is the format string (and dies if fails to find exactly one such arg). 2) with the new embed.fnc functionally, more functions have been marked with the 'f' flag. When such a function passes its fmt arg onto an inner printf-like function, we simply disable the warning for that call using GCC_DIAG_IGNORE(-Wformat-nonliteral), since we know that the caller must have already checked it. 3) In quite a few places the format string isn't literal, but it *is* constant (e.g. PL_warn_uninit_sv). For those cases, again disable the warning. 4) In pp_formline(), a particular format was was one of several different literal strings depending on circumstances. Rather than assigning this string to a temporary variable, incorporate the ?: branches directly in the function call arg. gcc is clever enough to decide the arg is then always literal.
* mark Perl_my_strftime with format attributeDavid Mitchell2013-11-281-3/+5
| | | | | | | | | | | | | | mark this function with __attribute__format__null_ok__(__strftime__,pTHX_1,0) so that compiler checks and warnings about strftime-style format args can be checked. Rather than adding new flag(s) to embed.fnc, I just enhanced the f flag to treat it as strftime-style rather than printf if the function name matches /strftime/. This was quicker, and we're unlikely to have many such functions.
* remove almost unreachable NULL sv arg code from sv_2*n_flagsDaniel Dragan2013-11-281-7/+7
| | | | | | | | | | | The NULL sv code being removed dates to commit e334a159a5 Perl 1.0 as the pre-SV str_2ptr and str_2num calls. When SVs were intoduced in commit 79072805bf Perl 5.0 alpha 2, the NULL sv code was copied to the new SV functions. The functions were bulk marked non-NULL in commit f54cb97a39 during 5.9.3 development. The docs were corrected to say NULLOK support in commit 53e8571218 during 5.11.0. See the perldelta part of this patch for the rest of commit body.
* mg.c: Extract code into a function.Karl Williamson2013-11-261-0/+1
| | | | | This is in preparation for the same code to be used in additional places. There should be no logic changes.
* Avoid pointer churn in study_chunk recursion bitmap allocationYves Orton2013-11-241-1/+1
| | | | | | | | | | | | | | | | Since we can only recurse into a given paren (or the entire pattern) once, we know that the maximum recursion depth is the number of parens in the pattern (plus one for "whole pattern"). This means we can preallocate one large bitmap, and then use different chunks of it for each level. That avoids SAVEFREEPV costs for each bitmap, which are likely short anyway. (One could imagine an optimization where a flag somewhere lets us use the RExC_study_chunk_recursed pointer as a bitmap, so we dont have to allocate all when we have less than 32 parens.) This removes the "recursed" argument from study_chunk() and replaces it with a "recursive_depth" argument which counts how deep we are in the bitmap "stack".
* [perl #120463] s/// and tr/// with wide delimitersFather Chrysostomos2013-11-141-1/+2
| | | | | | | | | | | | | | | | | | | | | $ perl -Mutf8 -e 's αaαα' Substitution replacement not terminated at -e line 1. What is happening is that the first scan goes past the delimiter at the end of the pattern. Then a single byte is compared (the previous character against the first byte of the opening delimiter) to see whether the parser needs to step back one byte before scanning the second part. That means you can do the equivalent of s/foo/|bar|g if / is replaced with a wide character: $ perl -l -Mutf8 -e '$_ = "a"; s αaα|b|; print' b This commit fixes it by giving toke.c:S_scan_str an extra parameter, so it can tell the callers that need this (scan_subst and scan_trans) where to start scanning the replacement.
* fix multi-eval of Perl_custom_op_xop in XopENTRYDaniel Dragan2013-11-101-1/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Commit 1830b3d9c8 introduced a flaw where XopENTRY calls Perl_custom_op_xop twice to retrieve the same XOP *. This is inefficient and causes extra machine code. Since I found no CPAN or upstream=blead usage of Perl_custom_op_xop, and its previous docs say it isn't 100% public, it is being converted to a macro. Most usage of Perl_custom_op_xop is to conditionally fetch a member of the XOP struct, which was previously implemented by XopENTRY. Move the XopENTRY logic and picking defaults to an expanded version of Perl_custom_op_xop. The union allows Perl_custom_op_get_field to return its result in 1 register, since the union is similar to a void * or IV, but with the machine code overhead of casting, if any, being done in the callee (Perl_custom_op_get_field), not the caller. Perl_custom_op_get_field can also return the XOP * without looking inside it to implement Perl_custom_op_xop. XopENTRYCUSTOM is a wrapper around Perl_custom_op_get_field with XopENTRY-like usage. XopENTRY is used by the OP_* macros, which are heavily used (but rarely called, since custom ops are rare) by Perl lang warnings system. The vararg warning arguments are usually evaluted no matter if the warning will be printed to STDERR or not. Since some people like to ignore warnings or run no strict; and warnings branches are frequent in pp_*, it is beneficial to make the OP_* macros smaller in machine code. The design of Perl_custom_op_get_field supports these goals. This commit does not pass judgement on Ben Morrow's unclear public or private API designation of Perl_custom_op_xop, and whether Perl_custom_op_xop should deprecated and removed from public API. It was trivial to leave a form of Perl_custom_op_xop in the new design. XOPe enums are identical to XOPf constants so no conversion has to be done between the field selector parameter and the field flag to test in machine code. ASSUME and NOT_REACHED are being introduced. The closest to the 2 previously was "assert(0)". Perl has not used ASSUME or CC specific versions of it before. Clang, GCC >= 4.5, and Visual C are supported. For completeness, ARMCC's __promise was added, but Perl is not known to have any support for ARMCC by this commiter. This patch is part of perl #115032.
* Fix qx, `` and <<`` overridesFather Chrysostomos2013-11-061-1/+0
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This resolves two RT tickets: • #115330 is that qx and `` overrides do not support interpolation. • #119827 is that <<`` does not support readpipe overrides at all. The obvious fix for #115330 fixes #119827 at the same time. When quote-like operators are parsed, after the string has been scanned S_sublex_push is called, which decides which of two paths to follow: 1) For things not requiring interpolation, the string is passed to tokeq (originally called q, it handles double backslashes and back- slashed delimiters) and returned to the parser immediately. 2) For anything that interpolates, the lexer enters a special inter- polation mode (LEX_INTERPPUSH) and goes through a more complex sequence over the next few calls (e.g., qq"a.$b.c" is turned into ‘stringify ( "a." . $ b . ".c" )’). When commit e3f73d4ed (Oct 2006, perl 5.10) added support for overrid- ing `` and qx with a readpipe sub, it did so by creating an entersub op in toke.c and making S_sublex_push follow path no. 1, taking the result if tokeq and inserting it into the already-constructed op tree for the sub call. That approach caused interpolation to be skipped when qx or `` is overridden. Furthermore it didn’t touch <<`` at all. The easiest solution is to let toke.c follow its normal path and create a backtick op (instead of trying to half-intercept it), and to deal with override lookup afterwards in ck_backtick, the same way require overrides are handled. Since <<`` also turns into a backtick op, it gets handled too that way.
* Put common override code into gv_overrideFather Chrysostomos2013-11-061-0/+2
| | | | | | | | | | | | When I moved the three occurrences of this code in op.c into a static function, I did not realise at the time that it also occurred thre etimes in toke.c. So now it is in a new non-static function in gv.c. Only two of the instances in toke.c could be changed to use this func- tion, as the otherwise is a little different. I couldn’t see a simple way of factoring its requirements in.
* Move the function to set $^X to its own filePeter Martini2013-11-041-0/+2
| | | | | | | This also moves the indirect dependency on stdbool.h to its own file, rather than being pulled in for all of perl.c, for those cases where one may want to test using other definitions of bool.
* Make mro code pass precomputed hash valuesFather Chrysostomos2013-11-041-1/+2
| | | | | | | | where possible This involved adding hv_fetchhek and hv_storehek macros and changing S_mro_clean_isarev to accept a hash parameter and expect HVhek_UTF8 instead of SVf_UTF8.
* [perl #119797] Fix if/else in lvalue subFather Chrysostomos2013-10-231-1/+2
| | | | | | | | | | | | | | | | | When if/else/unless is the last thing in an lvalue sub, the lvalue context is not always propagated properly and scope exit tries to copy things, including arrays, resulting in ‘Bizarre copy of ARRAY’. This commit fixes the bizarre copy by flagging any leave op that is part of an lvalue sub’s return sequence, using the OPpLEAVE flag added for this purpose in the previous commit. Then pp_leave uses that flag to avoid copying return values, but protects them via the mortals stack just like pp_leavesublv (actually pp_ctl.c:S_return_lvalues). For ‘if’ and ‘unless’ without ‘else’, the lvalue context was not being propagated, resulting in arrays’ getting flattened despite the lvalue context. op_lvalue_flags in op.c needed to handle AND and OR ops, which ‘if’ and ‘unless’ compile to, to make this work.
* make sv_2bool_flags() non-recursive on overloadDaniel Dragan2013-10-211-1/+1
| | | | | | | | | | | | | When Perl_sv_2bool_flags() has an overloaded arg, it calls SvTRUE() on the SV returned from the overload method. This indirectly calls sv_2bool_flags() again. Change it so that sv_2bool_flags() just iterates the new overload value each time. 2 callsites were converted to gotos. A SvTRUE_common was expanded so goto can be used. This function's machine code size on VC2003 32 bits dropped by 0x24 bytes after this patch.
* Adding a prototype attribute.Peter Martini2013-10-161-1/+2
| | | | | | | | | | | | | | | This attribute adds an additional way of declaring a prototype for a sub, making sub foo($$) and sub foo : prototype($$) equivalent. The intent is to keep the functionality of prototypes while allowing other modules to use the syntactic space it currently occupies for other purposes. The attribute is supported in attributes.xs to allow attributes::->import to work, but if its defined inline via something like sub foo : prototype($$) {}, it will not call out to the attributes module. For: RT #119251
* regcomp.c: Remove unused parameter in static functionKarl Williamson2013-09-241-2/+1
| | | | | This parameter is no longer used, since a few commits ago in this series.
* Teach regex optimizer to handle above-Latin1Karl Williamson2013-09-241-4/+4
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Until this commit, the regular expression optimizer has essentially punted on above-Latin1 code points. Under some circumstances, they would be taken into account, more or less, but often, the generated synthetic start class would end up matching all above-Latin1 code points. With the advent of inversion lists, it becomes feasible to actually fully handle such code points, as inversion lists are a convenient way to express arbitrary lists of code points and take their union, intersection, etc. This commit changes the optimizer to use inversion lists for operating on the code points the synthetic start class can match. I don't much understand the overall operation of the optimizer. I'm told that previous porters found that perturbing it caused unexpected behaviors. I had promised to get this change in 5.18, but didn't. I'm trying to get it in early enough into the 5.20 preliminary series that any problems will surface before 5.20 ships. This commit doesn't change the macro level logic, but does significantly change various micro level things. Thus the 'and' and 'or' subroutines have been rewritten to use inversion lists. I'm pretty confident that they do what their names suggest. I re-derived the equations for what these operations should do, getting the same results in some cases, but extending others where the previous code mostly punted. The derivations are given in comments in the respective routines. Some of the code is greatly simplified, as it no longer has to treat above-Latin1 specially. It is now feasible for /i matching of above-Latin1 code points to know explicitly the folds that should be in the synthetic start class. But more prepatory work needs to be done before putting that into place. ...
* regcomp.c: Add some static functionsKarl Williamson2013-09-241-0/+17
| | | | | | | | | This commit adds some functions that are currently unused, but will be used in a future commit. This commit is essentially to make the differences smaller in that commit, as 'diff' is getting confused and not outputting the logical differences. The functions are added in a block at the beginning of the file to avoid the 'diff' issues. A later white-space only commit will move them to more appropriate positions.
* regcomp.c: Use STR_WITH_LEN to avoid bookkeepingKarl Williamson2013-09-241-1/+1
| | | | | | By changing the order of the parameters to the static function S_add_data, we can call it with STR_WITH_LEN and avoid a human having to count characters.
* regcomp.c: Change some static parameters to constKarl Williamson2013-09-241-2/+2
| | | | I found I needed const in a future commit.
* regcomp.c: Add parameter to static functionKarl Williamson2013-09-241-1/+2
| | | | | | | This parameter will be used in future commits. This commit is really only to make the difference listing smaller in those, by committing separately just the book-keeping parts. This parameter requires also passing the aTHX_ thread parameter
* Make typedef fully typedefKarl Williamson2013-09-241-35/+36
| | | | | | | | | | | | | | | | | | | | | | The regcomp.c struct RExC_state_t has not been usable fully as a typedef, requiring the 'struct' at times. This has caused me, and I presume others, wasted time when we forget to use it under those circumstances when it should be used, but it's never been a big enough issue to cause me to spend tuits on it. But, working on something else, I finally came to the realization of what the problem is. It is because proto.h is #included before regcomp.h is, and so functions that are declared in proto.h that have something that is a RExC_state_t as a parameter don't know that it is a typedef because that is defined in regcomp.h. A way around this is already used for other similar structures, and that is to declare them in perl.h which is always read in before proto.h, leaving the definitions to regcomp.h. Thus proto.h knows enough to compile. The structure was already declared in perl.h; just not typedef'd. Otherwise proto.h would not know about it at all. This patch moves two regcomp.c related declarations in perl.h to the same section as the others, and changes the one for RExC_state_t to be a typedef. All the 'struct' uses are removed.
* regcomp.c: Change names of some static functionsKarl Williamson2013-09-241-8/+8
| | | | | | | | | | | | | | | | | The term 'class' is very overloaded in regex code and documentation. perlrecharclass.pod calls the dot (matching any char) a class, and calls the [] form "bracketed character classes". There are other meanings as well. This is the first commit in a short series that removes some of those overloadings. One instance of class is the "synthetic start class", generated by the regex optimizer to be a list of all the code points a sucessful match could possibly start with. This is useful in more quickly finding where to start looking in matching against a target string. Prior to this commit, the routines that referred to this began with 'cl_', and the formal parameters were 'cl', which could mean any class. This commit changes those instances of 'cl' to 'ssc' to indicate this is the only type of class that is being handled.
* regcomp.c: Rework static function call; commentsKarl Williamson2013-09-241-3/+2
| | | | | | The previous commit just extracted out code into a function. This commit renames a parameter for clarity, combines two parameters to make the interface cleaner, and adds and moves comments around.
* regcomp.c: Extract code into separate functionKarl Williamson2013-09-241-0/+7
| | | | | | A future commit will use this functionality from a second place. For now, just cut and paste, and do the minimal ancillary work to get it to compile and pass.
* Add regnode struct for synthetic start classKarl Williamson2013-09-241-8/+8
| | | | | | | | | | | | As part of extending the regular expression optimizer to properly handle above Latin1 code points, I need an inversion list to contain which code points the synthetic start class (ssc) matches. The ssc currently is the same as a locale-aware ANYOF node, which uses the struct of a regular ANYOF node, plus some extra fields at the end. This commit creates a new typedef for ssc use, which is the locale-aware ANYOF node, plus an extra SV* at the end to hold the inversion list.
* regcomp.c: Extract code into separate functionKarl Williamson2013-09-241-0/+1
| | | | | This is in preparation for it to be called from more than one place, in a future commit.
* Removed DUMP_FDS and dump_fds()Brian Fraser2013-09-211-3/+0
| | | | | | If perl was compiled with -DDUMP_FDS, it would define dump_fds and add it to the API, although even then nothing used it. dump_fds() itself was buggy, only checking for fds 0 through 32.
* toke.c, S_scan_ident(): Don't take a "end of buffer" argument, use PL_bufendBrian Fraser2013-09-181-1/+1
| | | | | | | | | | | | | | | All but one of scan_ident()'s callers already passed PL_bufend as the removed argument; The one deviant was intuit_more(), which was setting the "end of buffer" argument, to the next close-bracket. This commit modifies intuit_more() to temporarily set PL_bufend and then restore it. This was done as groundwork for the following commit, which will add more uses of PEEKSPACE() to scan_ident() in order to fix some whitespace and line number bugs, and PEEKSPACE() modifies PL_bufend directly if it encounters a newline at the end of the buffer -- that last bit being why changing intuit_more() to modify-and-restore PL_bufend is safe, since the end of the buffer will always be a ']'
* [perl #115928] a consistent (public) rand() implementationTony Cook2013-09-131-0/+2
| | | | | | | | | | | | | | | | Based on Yves's random branch work. This version makes the new random number visible to external modules, for example, List::Util's XS shuffle() implementation. I've also added a 64-bit implementation when HAS_QUAD is true, this should be significantly faster, even on 32-bit CPUs. This is intended to produce exactly the same sequence as the original implementation. The original version of this commit retained the "freebsd" name from Yves's original work for the function and data structure names. I've removed "freebsd" from most function names so the name isn't an issue if we choose to replace the implementation,
* perlapi: Add doc for my_strlcpy, my_strlcatKarl Williamson2013-09-101-2/+2
|
* gv.c: Split part of find_default_stash into gv_is_in_main.Brian Fraser2013-09-111-0/+2
| | | | | gv_is_in_main() checks if an unqualified identifier is in the main:: stash.
* gv.c: Rename magicalize_gv into gv_magicalize, make it more specific.Brian Fraser2013-09-111-1/+1
| | | | | Namely, gv_magicalize no longer stores the GV into the stash, which is gv_fetchpvn_flags' job.
* gv.c, gv_fetchpvn_flags: Split another chunk of magic-checking code.Brian Fraser2013-09-111-0/+1
| | | | | | This bit is called when a GV already exists, but it's name is length-one and it's on the main:: stash, so it might have multiple kinds of magic, like $! and %!, or @+ and %+.
* gv.c: Move the code that magicalizes new globs into magicalize_gv().Brian Fraser2013-09-111-0/+3
|
* gv.c: Begin splitting gv_fetchpvn_flags into smaller helper functions.Brian Fraser2013-09-111-0/+7
| | | | | This commit takes a chunk of code out of gv_fetchpvn_flags and turns it into two fuctions: parse_gv_stash_name and find_default_stash.
* regcomp.c: Make all warnings and error messages UTF-8 cleanBrian Fraser2013-09-101-1/+1
|
* [perl #117265] correctly handle overloaded stringsTony Cook2013-09-091-1/+1
|
* Revert "Let av_push accept NULL values"Father Chrysostomos2013-09-071-1/+1
| | | | | | | | | | | This reverts commit 7b6e8075e45ebc684565efbe3ce7b70435f20c79. It turns out to be problematic, because it causes NULLs on the stack, which XSUBs may trip on. My main reason for it was actually to try to resolve some CPAN failures, but it turns out that other fixes have removed the need for that.
* Fix PerlIO_get_cnt and friendsLeon Timmermans2013-09-071-4/+4
| | | | These functions worked with ints instead of SSize_t,
* Let av_push accept NULL valuesFather Chrysostomos2013-09-061-1/+1
| | | | | | | Now that NULL is used for a nonexistent element, it is easy for XS code to pass it to av_push(). av_store already accepts NULL, and av_push already works with it on non-debugging builds, so there is really no need for this restriction.
* Put AV defelem creation code in one placeFather Chrysostomos2013-09-061-0/+1
|
* [perl #115768] improve (caller)[2] line numbersFather Chrysostomos2013-09-011-1/+2
| | | | | | | | | | | | | warn and die have special code (closest_cop) to find a nulled nextstate op closest to the warn or die op, to get the line number from it. This commit extends that capability to caller, so that if (1) { foo(); } sub foo { warn +(caller)[2] } shows the right line number.
* utf8.c: Remove wrapper functions.Karl Williamson2013-08-291-13/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Now that the Unicode data is stored in native character set order, it is rare to need to work with the Unicode order. Traditionally, the real work was done in functions that worked with the Unicode order, and wrapper functions (or macros) were used to translate to/from native. There are two groups of functions: one that translates from code point to UTF-8, and the other group goes the opposite direction. This commit changes the base function that translates from UTF-8 to code point to output native instead of Unicode. Those extremely rare instances where Unicode output is needed instead will have to hand-wrap calls to this function with a translation macro, as now described in the API pod. Prior to this, it was the other way, the native was wrapped, and the rare, strict Unicode wasn't. This eliminates a layer of function call overhead for a common case. The base function that translates from code point to UTF-8 retains its Unicode input, as that is more natural to process. However, it is de-emphasized in the pod, with the functionality description moved to the pod for a native input wrapper function. And, those wrappers are now macros in all cases; previously there was function call overhead sometimes. (Equivalent exported functions are retained, however, for XS code that uses the Perl_foo() form.) I had hoped to rebase this commit, squashing it with an earlier commit in this series, eliminating the use of a temporary function name change, but the work involved turns out to be large, with no real payoff.
* utf8.c: Stop using two functionsKarl Williamson2013-08-291-2/+4
| | | | | | | | | | | | | | | | | This is in preparation for deprecating these functions, to force any code that has been using these functions to change. Since the Unicode tables are now stored in native order, these functions should only rarely be needed. However, the functionality of these is needed, and in actuality, on ASCII platforms, the native functions are #defined to these. So what this commit does is rename the functions to something else, and create wrappers with the old names, so that anyone using them will get the deprecation when it actually goes into effect: we are waiting for CPAN files distributed with the core to change before doing the deprecation. According to cpan.grep.me, this should affect fewer than 10 additional CPAN distributions.
* Convert uvuni_to_utf8() to functionKarl Williamson2013-08-291-1/+1
| | | | | | | Code should almost never be dealing with non-native code points This is in preparation for later deprecation when our CPAN modules have been converted away from using it.
* Deprecate utf8_to_uni_buf()Karl Williamson2013-08-291-1/+1
| | | | | | | Now that the tables are stored in native order, there is almost no need for code to be dealing in Unicode order. According to grep.cpan.me, there are no uses of this function in CPAN.
* Deprecate valid_utf8_to_uvuni()Karl Williamson2013-08-291-1/+1
| | | | | | | | | Now that all the tables are stored in native format, there is very little reason to use this function; and those who do need this kind of functionality should be using the bottom level routine, so as to make it clear they are doing nonstandard stuff. According to grep.cpan.me, there are no uses of this function in CPAN.
* utf8.c: Swap which fcn wraps the otherKarl Williamson2013-08-291-2/+1
| | | | This is in preparation for the current wrapee becoming deprecated
* Deprecate NATIVE_TO_NEED and ASCII_TO_NEEDKarl Williamson2013-08-291-0/+2
| | | | | | | | | | | | | | | | | | These macros are no longer called in the Perl core. This commit turns them into functions so that they can use gcc's deprecation facility. I believe these were defective right from the beginning, and I have struggled to understand what's going on. From the name, it appears NATIVE_TO_NEED taks a native byte and turns it into UTF-8 if the appropriate parameter indicates that. But that is impossible to do correctly from that API, as for variant characters, it needs to return two bytes. It could only work correctly if ch is an I8 byte, which isn't native, and hence the name would be wrong. Similar arguments for ASCII_TO_NEED. The function S_append_utf8_from_native_byte(const U8 byte, U8** dest) does what I think NATIVE_TO_NEED intended.
View Lorry Controller Status