gmane.emacs.devel

Re: problem report #101

Kenichi Handa — 2008-12-02T02:16:02

[...] Here, "face" is a return value of face_at_string/buffer_position, and thus is never NULL. Which do you think is better; ignore this waring or use a new macro FACE_FROM_ID_NO_CHECK instead of FACE_FROM_ID? /* Like FACE_FROM_ID but don't check the validity of ID. */ #define FACE_FROM_ID_NO_CHECK(F, ID)\ (FRAME_FACE_CACHE (F)->faces_by_id[ID]) --- Kenichi Handa handa< at >ni.aist.go.jp

Re: hash-table-{to, from}-alist

Andreas Schwab — 2008-12-02T00:13:49

That should be fairly easy by introducing a variable, say, hash-table-reader-function and just calling its value when set (and bailing out if not). That can then be implemented fully in Lisp. Andreas.

Re: multi-threaded Emacs

2008-12-01T22:41:07

Beautifully put.

Re: hash-table-{to, from}-alist

Ted Zlatanov — 2008-12-01T22:01:42

On Wed, 26 Nov 2008 15:16:41 -0600 Ted Zlatanov lifelogs.com> wrote: TZ> The patch below does all the previously discussed formatting plus TZ> detection of circular references and Davis Herring's suggestion of TZ> skipping the first preliminary space. The old-style hashtable printout TZ> is disabled with an #ifdef, so only the new style is available. The attached patch does reading and writing of hashtables. I believe it works correctly and handles the common error cases. Please review. Here are the issues: - I don't know how to manage a dynamic array of Lisp_Object objects, as make-hash-table wants. What I have right now is hacky and breaks for more than 5 parameters (e.g. reading #s(hash-table size 4 size 5 size 4 size 5 size 4 weakness x) will ignore the weakness parameter. Ideally I'd put the arguments in a list and pass that off to make-hash-table, but I couldn't find an example of that in the source (casting a list of an array of Lisp_Object elements). Additionally I use xmalloc without freeing the memory; I don't know if that will cause issues but I saw it used in other places without freeing the memory. - it seems to me that EQ(head, Qhash_table_size_marker) || EQ(head, Qhash_table_test_marker) || EQ(head, Qhash_table_weakness_marker) || EQ(head, Qhash_table_rehash_size_marker) || EQ(head, Qhash_table_rehash_threshold_marker)) can be written better, but I don't know enough to do it right. In addition, I don't know how to convert all those parameters to their : plist version, e.g. size => :size. I'm sure it's trivial but couldn't find it in the source. I really wish I could write all this in ELisp :) - my code is very deliberate, using CAR_SAFE and CDR_SAFE a lot. If that causes performance issues, please suggest improvements. Ditto for anything else--remember this is my first trip to the C side of Emacs and treat everything with suspicion. Thanks Ted

Re: hash-table-{to, from}-alist

Ted Zlatanov — 2008-12-01T21:53:02

On Thu, 27 Nov 2008 09:24:34 +0900 "Stephen J. Turnbull" xemacs.org> wrote: SJT> Ted Zlatanov writes: SJT> XEmacs abbreviates all "large" atomic objects in Lisp backtraces, SJT> AFAIK. This includes anything with a plist. But you can have your SJT> preference by setting `print-readably' globally. SM> char tables are always printed in full and we haven't had any complaints SM> about that, so you're probably right: let's just always print them SM> in full. If it turns out to be problematic in some cases, we can later SM> on add some extra variable to control the level of detail. SJT> We already have that. (^^; Can you and Stefan figure out whether to have print-readably or not? I was against it, and Stefan supported that. But let's see the pro argument and discuss this. I'd like to make your life easier in addition to providing features for GNU Emacs. Will the current state of things cause XEmacs breakage anywhere? One idea I had is to check if print-readably is bound; if not then we act as if it was bound and t. That will DTRT in GNU Emacs and in XEmacs, I hope. Thanks Ted

Re: multi-threaded Emacs

Stefan Monnier — 2008-12-01T20:34:21

Actually, no: while Emacs is waiting for socket data, it can already run other async code, such as process filters. So we can solve most I/O bound problems just by using a very coarse notion of concurrency. We already partly do it, except we're currently limited to a single stack, which makes coding painful and prevents some useful cases. Providing a cooperative threading model will remove this problem. But, no it's not a "final solution": I hope there never will be a final solution, because Emacs will always need to evolve. Stefan

Re: Making "problem report #FOO" mails easier to scan.

Reiner Steib — 2008-12-01T20:33:44

HTH... --8<---------------cut here---------------start------------->8--- (defvar rs-message-coverity-header "== Problem report from coverity, see end of message for more information ==") (defvar rs-message-coverity-footer "---------- Footnote: This problem report was created by scan.coverity.com. Anyone can register and read the reports, ...") (defun rs-message-coverity () "Insert suitable Subject, header and footer for a Coverity report. This command should be called from a message buffer which must contain the content of the report, including the pseudo header (CID, Function, ...)." (interactive) (save-excursion (message-goto-body) (when rs-message-coverity-header (insert rs-message-coverity-header) (insert "\n")) (re-search-forward "\ CID: \$[0-9]+\$ Checker: [^\n]* File: [^\n]* Function: \$[^\n]*\$ Description: \$[^\n]*\$" nil t) (let ( ;; (cid (match-string 1)) (fct (match-string 2)) (des (match-string 3))) (message-replace-header "Subject" (format "PR #%s: %s; %s" cid fct des)) (message-replace-header "To" "emacs-devel< at >gnu.org")) (when rs-message-coverity-footer (message-goto-signature) (unless (eobp) (forward-line -1)) (insert rs-message-coverity-footer) (insert "\n")))) ;; (define-key message-mode-map (kbd "") 'rs-message-coverity) --8<---------------cut here---------------end--------------->8--- Bye, Reiner.

Re: vc-dir header for bzr

Karl Fogel — 2008-12-01T20:30:13

I'm actually not much of a bzr user yet. However, I'm becoming one soon, thanks to a fortuitous intersection of work and interest. So I'll try to remember to look at this once I'm more expert -- but feel free to poke me again in month or so too. -Karl

Re: problem report #104

Dan Nicolaescu — 2008-12-01T20:29:50

> Dan Nicolaescu ics.uci.edu> writes: > > > Event returned_null: Function "malloc" returned NULL value (checked 29 out of 36 times) > > In practice, this never happens. True, but xmalloc is available in that file, so I just made it use xmalloc.

X window display bug?

Bo Lin — 2008-12-01T14:36:34

Hi emacs-devel, I'm getting weird GUI behavior when compiling the latest CVS code (actually I use git to pull from git://git.sv.gnu.org/emacs.git). Attached is a screen shot of what it (started with emacs -Q) looks like. As you can see, the left and right margins are unusually large, I can't `split-window-horizontally' because "Window width 7 too small (after splitting)", and the number returned by `window-width' is clearly wrong. This behavior has been around for at least two months (it has been like this ever since I started trying to compile emacs23). Any suggestions? This is compiled on Ubuntu 8.04. $ uname -a Linux unicorn 2.6.24-22-generic #1 SMP Mon Nov 24 18:32:42 UTC 2008 i686 GNU/Linux $ gcc -v Using built-in specs. Target: i486-linux-gnu Configured with: ../src/configure -v --enable-languages=c,c++,fortran,objc,obj-c++,treelang --prefix=/usr --enable-shared --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --enable-nls --with-gxx-include-dir=/usr/include/c++/4.2 --program-suffix=-4.2 --enable-clocale=gnu --enable-libstdcxx-debug --enable-objc-gc --enable-mpfr --enable-targets=all --enable-checking=release --build=i486-linux-gnu --host=i486-linux-gnu --target=i486-linux-gnu Thread model: posix gcc version 4.2.4 (Ubuntu 4.2.4-1ubuntu3)

Re: problem report #100

Stefan Monnier — 2008-12-01T20:25:02

This is indeed the best way to solve the non-bugs. It's usually better to make the correctness obvious than to add a comment explaining why this is correct. Stefan

Re: problem report #105

Andreas Schwab — 2008-12-01T20:14:49

Fixed. Andreas.

vc-dir header for bzr (was: Re: [PATCH] vc-bzr.el: avoid stompingfiles across hardlink branches.)

Dan Nicolaescu — 2008-12-01T20:10:34

Since you seem to be interested in VC with bzr, can you please look at the following issue too? In vc-dir it is possible to add custom headers for each VC backend. Currently for bzr only a single one is used "Parent branch". But it seems that "bzr info" offers more information that can be useful: $ bzr info Standalone tree (format: pack-0.92) Location: branch root: FOOBAR Related branches: push branch: FOO parent branch: BAR Is there anything else that could be added to the headers here? Thanks. --dan

Re: Making "problem report #FOO" mails easier to scan.

Karl Fogel — 2008-12-01T19:24:15

Thanks. I suspected they came from coverity, but there was no sign of that in the emails themselves. Maybe just insert a standard footer explaining it, if you have time to set that up. -Karl

Re: Making "problem report #FOO" mails easier to scan.

Dan Nicolaescu — 2008-12-01T19:16:07

> Would it be possible to make "problem report" mails have Subject headers > like this: It would, but it would mean more work for me... This is all done manually, copy and paste from a few places on a web page... If someone wants to write a gnus hook to fill in the Subject, I could use it. Or if someone else wants to do this, even better. > Also, is there a URL associated with each problem report? If so, it AFAICT no, the URL is associated with a login name and a session. > would be nice if the report included that URL. (Right now, one can't > tell from looking at one of the mails what is generating them.) Similar reports were discussed/resolved on the list a while ago. They come from scan.coverity.com. Anyone can register and read the reports, but so far nobody volunteered to do it...

Re: problem report #102

Dan Nicolaescu — 2008-12-01T19:00:56

> Dan Nicolaescu ics.uci.edu> writes: > > > Andreas Schwab suse.de> writes: > > > > > Dan Nicolaescu ics.uci.edu> writes: > > > > > > > Event negative_returns: Negative constant "-1" was passed to a negative sink. [model] > > > > > > What's a "negative sink"?? > > > > Thanks for looking at these. > > > > negative_sink: The checker reports this event when it determines that a value is used in an operation that would have a bad effect if that value were negative. The checker then tracks the value used to see if it is subsequently compared against a negative value. > > So what _is_ the "bad effect"? Sorry, no idea, I don't have any more information that you do. I didn't even look at the code, just copy+paste...

Re: multi-threaded Emacs

Giuseppe Scrivano — 2008-12-01T18:57:44

If we don't want to do immediately all changes at one time and proceed slowly to a real parallelism then we can start with a concurrent threads model. I think that concurrent threads can't be considered the final solution, what I would like to optimize using threads is I/O bound and a concurrent model can do it only in part. If an Elisp primitive is waiting for some data from a file or a socket, Emacs will hang until it is ready. This problem can't be solved by concurrent threads that will switch only inside the QUIT macro. If we will implement a concurrent threads model and introduce the QUIT macro after it can't be used by parallel threads as accesses to shared data must be synchronized differently. IMHO, concurrent threads itself are not very useful, they can be useful if considered a first step in the direction of a real parallelism. I know it is a lot of work and the question (as in everything) is: does benefits we will have worth this effort? Regards, Giuseppe

Re: problem report #102

Andreas Schwab — 2008-12-01T18:51:45

So what _is_ the "bad effect"? I could not find any indication of an alleged "bad effect". Andreas.

Making "problem report #FOO" mails easier to scan.

Karl Fogel — 2008-12-01T18:48:21

Would it be possible to make "problem report" mails have Subject headers like this: Subject: problem report #101: Variable "face" tracked as NULL [...] ? Or maybe even shorten the prefix, so more of the Description string is visible: Subject: PR #101: Variable "face" tracked as NULL was passed to [...] ? I assume the entire description string would be included. Above, I'm just eliding it the way a mailreader might cut it off along the right edge -- say, as a Gnus Summary buffer would. It's time-consuming to have to open the mail to get an idea of what the problem report is about. All the specific content is in the body: From: Dan Nicolaescu ics.uci.edu> Subject: problem report #101 To: emacs-devel< at >gnu.org Date: Mon, 1 Dec 2008 08:59:25 -0800 (PST) CID: 101 Checker: FORWARD_NULL (help) File: base/src/emacs/src/font.c Function: font_at Description: Variable "face" tracked as NULL was passed to a function that dereferences it. Also, is there a URL associated with each problem report? If so, it would be nice if the report included that URL. (Right now, one can't tell from looking at one of the mails what is generating them.) -Karl

Re: problem report #102

Dan Nicolaescu — 2008-12-01T18:34:46

> Dan Nicolaescu ics.uci.edu> writes: > > > Event negative_returns: Negative constant "-1" was passed to a negative sink. [model] > > What's a "negative sink"?? Thanks for looking at these. negative_sink: The checker reports this event when it determines that a value is used in an operation that would have a bad effect if that value were negative. The checker then tracks the value used to see if it is subsequently compared against a negative value. [model] are links to other pages with further explanations for the checker's action. Following those links: At conditional (1): "s != 0" taking true path 1605 if (s) 1606 s[0] = '\1'; 1607 Event neg_sink_parm_call: Parameter "maxsize" passed to negative sink "emacs_strftimeu" [model] 1608 result = emacs_strftimeu (s, maxsize, format, tp, ut); 1609 1484 size_t 1485 emacs_strftimeu (s, maxsize, format, tp, ut) 1486 char *s; 1487 size_t maxsize; 1488 const char *format; 1489 const struct tm *tp; 1490 int ut; 1491 { Event neg_sink_parm_call: Parameter "maxsize" passed to negative sink "nstrftime" [model] 1492 return my_strftime (s, maxsize, format, tp, ut, 0); 1493 } 1494 #endif 521 /* Write information from TP into S according to the format 522 string FORMAT, writing no more that MAXSIZE characters 523 (including the terminating '\0') and returning number of 524 characters written. If S is NULL, nothing will be written 525 anywhere, so to determine how many characters would be 526 written, use NULL for S and (size_t) UINT_MAX for MAXSIZE. */ 527 size_t 528 my_strftime (s, maxsize, format, tp extra_args LOCALE_PARAM) 529 CHAR_T *s; Event sizet: size_t parameter "maxsize" detected 530 size_t maxsize; 531 const CHAR_T *format; 532 const struct tm *tp; 533 extra_args_spec 534 LOCALE_PARAM_DECL 535 { 536 #if defined _LIBC && defined USE_IN_EXTENDED_LOCALE_MODEL 537 struct locale_data *const current = loc->__locales[LC_TIME]; 538 #endif 539 540 int hour12 = tp->tm_hour; 541 #ifdef _NL_CURRENT 542 /* We cannot make the following values variables since we must delay 543 the evaluation of these values until really needed since some 544 expressions might not be valid in every situation. The `struct tm' 545 might be generated by a strptime() call that initialized 546 only a few elements. Dereference the pointers only if the format 547 requires this. Then it is ok to fail if the pointers are invalid. */ 548 # define a_wkday \ 549 ((const CHAR_T *) _NL_CURRENT (LC_TIME, NLW(ABDAY_1) + tp->tm_wday)) 550 # define f_wkday \ 551 ((const CHAR_T *) _NL_CURRENT (LC_TIME, NLW(DAY_1) + tp->tm_wday)) 552 # define a_month \ 553 ((const CHAR_T *) _NL_CURRENT (LC_TIME, NLW(ABMON_1) + tp->tm_mon)) 554 # define f_month \ 555 ((const CHAR_T *) _NL_CURRENT (LC_TIME, NLW(MON_1) + tp->tm_mon)) 556 # define ampm \ 557 ((const CHAR_T *) _NL_CURRENT (LC_TIME, tp->tm_hour > 11 \ 558 ? NLW(PM_STR) : NLW(AM_STR))) 559 560 # define aw_len STRLEN (a_wkday) 561 # define am_len STRLEN (a_month) 562 # define ap_len STRLEN (ampm) 563 #else 564 # if !HAVE_STRFTIME 565 # define f_wkday (weekday_name[tp->tm_wday]) 566 # define f_month (month_name[tp->tm_mon]) 567 # define a_wkday f_wkday 568 # define a_month f_month 569 # define ampm (L_("AMPM") + 2 * (tp->tm_hour > 11)) 570 571 size_t aw_len = 3; 572 size_t am_len = 3; 573 size_t ap_len = 2; 574 # endif 575 #endif 576 const char *zone; 577 size_t i = 0; 578 CHAR_T *p = s; 579 const CHAR_T *f; 580 #if DO_MULTIBYTE && !defined COMPILE_WIDE 581 const char *format_end = NULL; 582 #endif 583 584 zone = NULL; 585 #if HAVE_TM_ZONE 586 /* The POSIX test suite assumes that setting 587 the environment variable TZ to a new value before calling strftime() 588 will influence the result (the %Z format) even if the information in 589 TP is computed with a totally different time zone. 590 This is bogus: though POSIX allows bad behavior like this, 591 POSIX does not require it. Do the right thing instead. */ 592 zone = (const char *) tp->tm_zone; 593 #endif 594 #if HAVE_TZNAME 595 if (ut) 596 { 597 if (! (zone && *zone)) 598 zone = "GMT"; 599 } 600 else 601 { 602 /* POSIX.1 requires that local time zone information be used as 603 though strftime called tzset. */ 604 # if HAVE_TZSET 605 tzset (); 606 # endif 607 } 608 #endif 609 610 if (hour12 > 12) 611 hour12 -= 12; 612 else 613 if (hour12 == 0) 614 hour12 = 12; 615 616 for (f = format; *f != '\0'; ++f) 617 { 618 int pad = 0;/* Padding for number ('-', '_', or 0). */ 619 int modifier;/* Field modifier ('E', 'O', or 0). */ 620 int digits;/* Max digits for numeric format. */ 621 int number_value; /* Numeric value to be printed. */ 622 int negative_number;/* 1 if the number is negative. */ 623 const CHAR_T *subfmt; 624 CHAR_T *bufp; 625 CHAR_T buf[1 + (sizeof (int) < sizeof (time_t) 626 ? INT_STRLEN_BOUND (time_t) 627 : INT_STRLEN_BOUND (int))]; 628 int width = -1; 629 int to_lowcase = 0; 630 int to_uppcase = 0; 631 int change_case = 0; 632 int format_char; 633 634 #if DO_MULTIBYTE && !defined COMPILE_WIDE 635 switch (*f) 636 { 637 case L_('%'): 638 break; 639 640 case L_('\b'): case L_('\t'): case L_('\n'): 641 case L_('\v'): case L_('\f'): case L_('\r'): 642 case L_(' '): case L_('!'): case L_('"'): case L_('#'): case L_('&'): 643 case L_('\''): case L_('('): case L_(')'): case L_('*'): case L_('+'): 644 case L_(','): case L_('-'): case L_('.'): case L_('/'): case L_('0'): 645 case L_('1'): case L_('2'): case L_('3'): case L_('4'): case L_('5'): 646 case L_('6'): case L_('7'): case L_('8'): case L_('9'): case L_(':'): 647 case L_(';'): case L_('<'): case L_('='): case L_('>'): case L_('?'): 648 case L_('A'): case L_('B'): case L_('C'): case L_('D'): case L_('E'): 649 case L_('F'): case L_('G'): case L_('H'): case L_('I'): case L_('J'): 650 case L_('K'): case L_('L'): case L_('M'): case L_('N'): case L_('O'): 651 case L_('P'): case L_('Q'): case L_('R'): case L_('S'): case L_('T'): 652 case L_('U'): case L_('V'): case L_('W'): case L_('X'): case L_('Y'): 653 case L_('Z'): case L_('['): case L_('\\'): case L_(']'): case L_('^'): 654 case L_('_'): case L_('a'): case L_('b'): case L_('c'): case L_('d'): 655 case L_('e'): case L_('f'): case L_('g'): case L_('h'): case L_('i'): 656 case L_('j'): case L_('k'): case L_('l'): case L_('m'): case L_('n'): 657 case L_('o'): case L_('p'): case L_('q'): case L_('r'): case L_('s'): 658 case L_('t'): case L_('u'): case L_('v'): case L_('w'): case L_('x'): 659 case L_('y'): case L_('z'): case L_('{'): case L_('|'): case L_('}'): 660 case L_('~'): 661 /* The C Standard requires these 98 characters (plus '%') to 662 be in the basic execution character set. None of these 663 characters can start a multibyte sequence, so they need 664 not be analyzed further. */ 665 add (1, *p = *f); 666 continue; 667 668 default: 669 /* Copy this multibyte sequence until we reach its end, find 670 an error, or come back to the initial shift state. */ 671 { 672 mbstate_t mbstate = mbstate_zero; 673 size_t len = 0; 674 size_t fsize; 675 676 if (! format_end) 677 format_end = f + strlen (f) + 1; 678 fsize = format_end - f; 679 680 do 681 { 682 size_t bytes = mbrlen (f + len, fsize - len, &mbstate); 683 684 if (bytes == 0) 685 break; 686 687 if (bytes == (size_t) -2) 688 { 689 len += strlen (f + len); 690 break; 691 } 692 693 if (bytes == (size_t) -1) 694 { 695 len++; 696 break; 697 } 698 699 len += bytes; 700 } 701 while (! mbsinit (&mbstate)); 702 703 cpy (len, f); 704 f += len - 1; 705 continue; 706 } 707 } 708 709 #else /* ! DO_MULTIBYTE */ 710 711 /* Either multibyte encodings are not supported, they are 712 safe for formats, so any non-'%' byte can be copied through, 713 or this is the wide character version. */ 714 if (*f != L_('%')) 715 { 716 add (1, *p = *f); 717 continue; 718 } 719 720 #endif /* ! DO_MULTIBYTE */ 721 722 /* Check for flags that can modify a format. */ 723 while (1) 724 { 725 switch (*++f) 726 { 727 /* This influences the number formats. */ 728 case L_('_'): 729 case L_('-'): 730 case L_('0'): 731 pad = *f; 732 continue; 733 734 /* This changes textual output. */ 735 case L_('^'): 736 to_uppcase = 1; 737 continue; 738 case L_('#'): 739 change_case = 1; 740 continue; 741 742 default: 743 break; 744 } 745 break; 746 } 747 748 /* As a GNU extension we allow to specify the field width. */ 749 if (ISDIGIT (*f)) 750 { 751 width = 0; 752 do 753 { At conditional (1): "width == 214748364" taking false path At conditional (3): "width > 214748364" taking true path At conditional (5): "width > 214748364" taking false path At conditional (6): "width == 214748364" taking false path At conditional (8): "width > 214748364" taking false path At conditional (9): "width == 214748364" taking true path At conditional (10): "(*f - 48) > 7" taking false path At conditional (12): "width > 214748364" taking true path 754 if (width > INT_MAX / 10 755 || (width == INT_MAX / 10 && *f - L_('0') > INT_MAX % 10)) 756 /* Avoid overflow. */ 757 width = INT_MAX; 758 else 759 { 760 width *= 10; 761 width += *f - L_('0'); 762 } 763 ++f; At conditional (2): "(*f - 48) <= 9" taking true path At conditional (4): "(*f - 48) <= 9" taking true path At conditional (7): "(*f - 48) <= 9" taking true path At conditional (11): "(*f - 48) <= 9" taking true path At conditional (13): "(*f - 48) <= 9" taking false path 764 } 765 while (ISDIGIT (*f)); 766 } 767 768 /* Check for modifiers. */ 769 switch (*f) 770 { At conditional (14): "*f == 69" taking true path 771 case L_('E'): 772 case L_('O'): 773 modifier = *f++; 774 break; 775 776 default: 777 modifier = 0; 778 break; 779 } 780 781 /* Now do the specified format. */ 782 format_char = *f; 783 switch (format_char) 784 { 785 #define DO_NUMBER(d, v) \ 786 digits = d > width ? d : width; \ 787 number_value = v; goto do_number 788 #define DO_NUMBER_SPACEPAD(d, v) \ 789 digits = d > width ? d : width; \ 790 number_value = v; goto do_number_spacepad 791 792 case L_('%'): 793 if (modifier != 0) 794 goto bad_format; 795 add (1, *p = *f); 796 break; 797 798 case L_('a'): 799 if (modifier != 0) 800 goto bad_format; 801 if (change_case) 802 { 803 to_uppcase = 1; 804 to_lowcase = 0; 805 } 806 #if defined _NL_CURRENT || !HAVE_STRFTIME 807 cpy (aw_len, a_wkday); 808 break; 809 #else 810 goto underlying_strftime; 811 #endif 812 813 case 'A': 814 if (modifier != 0) 815 goto bad_format; 816 if (change_case) 817 { 818 to_uppcase = 1; 819 to_lowcase = 0; 820 } 821 #if defined _NL_CURRENT || !HAVE_STRFTIME 822 cpy (STRLEN (f_wkday), f_wkday); 823 break; 824 #else 825 goto underlying_strftime; 826 #endif 827 828 case L_('b'): 829 case L_('h'): 830 if (change_case) 831 { 832 to_uppcase = 1; 833 to_lowcase = 0; 834 } 835 if (modifier != 0) 836 goto bad_format; 837 #if defined _NL_CURRENT || !HAVE_STRFTIME 838 cpy (am_len, a_month); 839 break; 840 #else 841 goto underlying_strftime; 842 #endif 843 844 case L_('B'): 845 if (modifier != 0) 846 goto bad_format; 847 if (change_case) 848 { 849 to_uppcase = 1; 850 to_lowcase = 0; 851 } 852 #if defined _NL_CURRENT || !HAVE_STRFTIME 853 cpy (STRLEN (f_month), f_month); 854 break; 855 #else 856 goto underlying_strftime; 857 #endif 858 859 case L_('c'): 860 if (modifier == L_('O')) 861 goto bad_format; 862 #ifdef _NL_CURRENT 863 if (! (modifier == 'E' 864 && (*(subfmt = 865 (const CHAR_T *) _NL_CURRENT (LC_TIME, 866 NLW(ERA_D_T_FMT))) 867 != '\0'))) 868 subfmt = (const CHAR_T *) _NL_CURRENT (LC_TIME, NLW(D_T_FMT)); 869 #else 870 # if HAVE_STRFTIME 871 goto underlying_strftime; 872 # else 873 subfmt = L_("%a %b %e %H:%M:%S %Y"); 874 # endif 875 #endif 876 877 subformat: 878 { 879 CHAR_T *old_start = p; 880 size_t len = my_strftime (NULL, (size_t) -1, subfmt, 881 tp extra_args LOCALE_ARG); 882 add (len, my_strftime (p, maxsize - i, subfmt, 883 tp extra_args LOCALE_ARG)); 884 885 if (to_uppcase) 886 while (old_start < p) 887 { 888 *old_start = TOUPPER ((UCHAR_T) *old_start, loc); 889 ++old_start; 890 } 891 } 892 break; 893 894 #if HAVE_STRFTIME && ! (defined _NL_CURRENT && HAVE_STRUCT_ERA_ENTRY) 895 underlying_strftime: 896 { 897 /* The relevant information is available only via the 898 underlying strftime implementation, so use that. */ 899 char ufmt[4]; 900 char *u = ufmt; 901 char ubuf[1024]; /* enough for any single format in practice */ 902 size_t len; 903 /* Make sure we're calling the actual underlying strftime. 904 In some cases, config.h contains something like 905 "#define strftime rpl_strftime". */ 906 # ifdef strftime 907 # undef strftime 908 size_t strftime (); 909 # endif 910 911 #ifdef STRFTIME_NO_POSIX2 912 /* Some system libraries do not support the POSIX.2 extensions. 913 In those cases, convert %h to %b, and strip modifiers. */ 914 modifier = 0; 915 if (format_char == 'h') 916 format_char = 'b'; 917 #endif 918 *u++ = '%'; 919 if (modifier != 0) 920 *u++ = modifier; 921 *u++ = format_char; 922 *u = '\0'; 923 len = strftime (ubuf, sizeof ubuf, ufmt, tp); 924 if (len == 0 && ubuf[0] != '\0') 925 return 0; 926 cpy (len, ubuf); 927 } 928 break; 929 #endif 930 931 case L_('C'): 932 if (modifier == L_('O')) 933 goto bad_format; 934 if (modifier == L_('E')) 935 { 936 #if HAVE_STRUCT_ERA_ENTRY 937 struct era_entry *era = _nl_get_era_entry (tp HELPER_LOCALE_ARG); 938 if (era) 939 { 940 # ifdef COMPILE_WIDE 941 size_t len = __wcslen (era->era_wname); 942 cpy (len, era->era_wname); 943 # else 944 size_t len = strlen (era->era_name); 945 cpy (len, era->era_name); 946 # endif 947 break; 948 } 949 #else 950 # if HAVE_STRFTIME 951 goto underlying_strftime; 952 # endif 953 #endif 954 } 955 956 { 957 int year = tp->tm_year + TM_YEAR_BASE; 958 DO_NUMBER (1, year / 100 - (year % 100 < 0)); 959 } 960 961 case L_('x'): 962 if (modifier == L_('O')) 963 goto bad_format; 964 #ifdef _NL_CURRENT 965 if (! (modifier == L_('E') 966 && (*(subfmt = 967 (const CHAR_T *)_NL_CURRENT (LC_TIME, NLW(ERA_D_FMT))) 968 != L_('\0')))) 969 subfmt = (const CHAR_T *) _NL_CURRENT (LC_TIME, NLW(D_FMT)); 970 goto subformat; 971 #else 972 # if HAVE_STRFTIME 973 goto underlying_strftime; 974 # else 975 /* Fall through. */ 976 # endif 977 #endif 978 case L_('D'): 979 if (modifier != 0) 980 goto bad_format; 981 subfmt = L_("%m/%d/%y"); 982 goto subformat; 983 984 case L_('d'): 985 if (modifier == L_('E')) 986 goto bad_format; 987 988 DO_NUMBER (2, tp->tm_mday); 989 990 case L_('e'): 991 if (modifier == L_('E')) 992 goto bad_format; 993 994 DO_NUMBER_SPACEPAD (2, tp->tm_mday); 995 996 /* All numeric formats set DIGITS and NUMBER_VALUE and then 997 jump to one of these two labels. */ 998 999 do_number_spacepad: 1000 /* Force `_' flag unless overridden by `0' or `-' flag. */ 1001 if (pad != L_('0') && pad != L_('-')) 1002 pad = L_('_'); 1003 1004 do_number: 1005 /* Format the number according to the MODIFIER flag. */ 1006 1007 if (modifier == L_('O') && 0 <= number_value) 1008 { 1009 #ifdef _NL_CURRENT 1010 /* Get the locale specific alternate representation of 1011 the number NUMBER_VALUE. If none exist NULL is returned. */ 1012 const CHAR_T *cp = nl_get_alt_digit (number_value 1013 HELPER_LOCALE_ARG); 1014 1015 if (cp != NULL) 1016 { 1017 size_t digitlen = STRLEN (cp); 1018 if (digitlen != 0) 1019 { 1020 cpy (digitlen, cp); 1021 break; 1022 } 1023 } 1024 #else 1025 # if HAVE_STRFTIME 1026 goto underlying_strftime; 1027 # endif 1028 #endif 1029 } 1030 { 1031 unsigned int u = number_value; 1032 1033 bufp = buf + sizeof (buf) / sizeof (buf[0]); 1034 negative_number = number_value < 0; 1035 1036 if (negative_number) 1037 u = -u; 1038 1039 do 1040 *--bufp = u % 10 + L_('0'); 1041 while ((u /= 10) != 0); 1042 } 1043 1044 do_number_sign_and_padding: 1045 if (negative_number) 1046 *--bufp = L_('-'); 1047 1048 if (pad != L_('-')) 1049 { 1050 int padding = digits - (buf + (sizeof (buf) / sizeof (buf[0])) 1051 - bufp); 1052 1053 if (padding > 0) 1054 { 1055 if (pad == L_('_')) 1056 { 1057 if ((size_t) padding >= maxsize - i) 1058 return 0; 1059 1060 if (p) 1061 memset_space (p, padding); 1062 i += padding; 1063 width = width > padding ? width - padding : 0; 1064 } 1065 else 1066 { 1067 if ((size_t) digits >= maxsize - i) 1068 return 0; 1069 1070 if (negative_number) 1071 { 1072 ++bufp; 1073 1074 if (p) 1075 *p++ = L_('-'); 1076 ++i; 1077 } 1078 1079 if (p) 1080 memset_zero (p, padding); 1081 i += padding; 1082 width = 0; 1083 } 1084 } 1085 } 1086 1087 cpy (buf + sizeof (buf) / sizeof (buf[0]) - bufp, bufp); 1088 break; 1089 1090 case L_('F'): 1091 if (modifier != 0) 1092 goto bad_format; 1093 subfmt = L_("%Y-%m-%d"); 1094 goto subformat; 1095 1096 case L_('H'): 1097 if (modifier == L_('E')) 1098 goto bad_format; 1099 1100 DO_NUMBER (2, tp->tm_hour); 1101 1102 case L_('I'): 1103 if (modifier == L_('E')) 1104 goto bad_format; 1105 1106 DO_NUMBER (2, hour12); 1107 1108 case L_('k'):/* GNU extension. */ 1109 if (modifier == L_('E')) 1110 goto bad_format; 1111 1112 DO_NUMBER_SPACEPAD (2, tp->tm_hour); 1113 1114 case L_('l'):/* GNU extension. */ 1115 if (modifier == L_('E')) 1116 goto bad_format; 1117 1118 DO_NUMBER_SPACEPAD (2, hour12); 1119 1120 case L_('j'): 1121 if (modifier == L_('E')) 1122 goto bad_format; 1123 1124 DO_NUMBER (3, 1 + tp->tm_yday); 1125 1126 case L_('M'): 1127 if (modifier == L_('E')) 1128 goto bad_format; 1129 1130 DO_NUMBER (2, tp->tm_min); 1131 1132 case L_('m'): 1133 if (modifier == L_('E')) 1134 goto bad_format; 1135 1136 DO_NUMBER (2, tp->tm_mon + 1); 1137 1138 #ifndef _LIBC 1139 case L_('N'):/* GNU extension. */ 1140 if (modifier == L_('E')) 1141 goto bad_format; 1142 1143 number_value = ns; 1144 if (width != -1) 1145 { 1146 /* Take an explicit width less than 9 as a precision. */ 1147 int j; 1148 for (j = width; j < 9; j++) 1149 number_value /= 10; 1150 } 1151 1152 DO_NUMBER (9, number_value); 1153 #endif 1154 1155 case L_('n'): 1156 add (1, *p = L_('\n')); 1157 break; 1158 1159 case L_('P'): 1160 to_lowcase = 1; 1161 #if !defined _NL_CURRENT && HAVE_STRFTIME 1162 format_char = L_('p'); 1163 #endif 1164 /* FALLTHROUGH */ 1165 1166 case L_('p'): 1167 if (change_case) 1168 { 1169 to_uppcase = 0; 1170 to_lowcase = 1; 1171 } 1172 #if defined _NL_CURRENT || !HAVE_STRFTIME 1173 cpy (ap_len, ampm); 1174 break; 1175 #else 1176 goto underlying_strftime; 1177 #endif 1178 1179 case L_('R'): 1180 subfmt = L_("%H:%M"); 1181 goto subformat; 1182 1183 case L_('r'): 1184 #ifdef _NL_CURRENT 1185 if (*(subfmt = (const CHAR_T *) _NL_CURRENT (LC_TIME, 1186 NLW(T_FMT_AMPM))) 1187 == L_('\0')) 1188 #endif 1189 subfmt = L_("%I:%M:%S %p"); 1190 goto subformat; 1191 1192 case L_('S'): 1193 if (modifier == L_('E')) 1194 goto bad_format; 1195 1196 DO_NUMBER (2, tp->tm_sec); 1197 1198 case L_('s'):/* GNU extension. */ 1199 { 1200 struct tm ltm; 1201 time_t t; 1202 1203 ltm = *tp; 1204 t = mktime (<m); 1205 1206 /* Generate string value for T using time_t arithmetic; 1207 this works even if sizeof (long) < sizeof (time_t). */ 1208 1209 bufp = buf + sizeof (buf) / sizeof (buf[0]); 1210 negative_number = t < 0; 1211 1212 do 1213 { 1214 int d = t % 10; 1215 t /= 10; 1216 1217 if (negative_number) 1218 { 1219 d = -d; 1220 1221 /* Adjust if division truncates to minus infinity. */ 1222 if (0 < -1 % 10 && d < 0) 1223 { 1224 t++; 1225 d += 10; 1226 } 1227 } 1228 1229 *--bufp = d + L_('0'); 1230 } 1231 while (t != 0); 1232 1233 digits = 1; 1234 goto do_number_sign_and_padding; 1235 } 1236 1237 case L_('X'): 1238 if (modifier == L_('O')) 1239 goto bad_format; 1240 #ifdef _NL_CURRENT 1241 if (! (modifier == L_('E') 1242 && (*(subfmt = 1243 (const CHAR_T *) _NL_CURRENT (LC_TIME, NLW(ERA_T_FMT))) 1244 != L_('\0')))) 1245 subfmt = (const CHAR_T *) _NL_CURRENT (LC_TIME, NLW(T_FMT)); 1246 goto subformat; 1247 #else 1248 # if HAVE_STRFTIME 1249 goto underlying_strftime; 1250 # else 1251 /* Fall through. */ 1252 # endif 1253 #endif 1254 case L_('T'): 1255 subfmt = L_("%H:%M:%S"); 1256 goto subformat; 1257 1258 case L_('t'): 1259 add (1, *p = L_('\t')); 1260 break; 1261 1262 case L_('u'): 1263 DO_NUMBER (1, (tp->tm_wday - 1 + 7) % 7 + 1); 1264 1265 case L_('U'): 1266 if (modifier == L_('E')) 1267 goto bad_format; 1268 1269 DO_NUMBER (2, (tp->tm_yday - tp->tm_wday + 7) / 7); 1270 1271 case L_('V'): 1272 case L_('g'): 1273 case L_('G'): 1274 if (modifier == L_('E')) 1275 goto bad_format; 1276 { 1277 int year = tp->tm_year + TM_YEAR_BASE; 1278 int days = iso_week_days (tp->tm_yday, tp->tm_wday); 1279 1280 if (days < 0) 1281 { 1282 /* This ISO week belongs to the previous year. */ 1283 year--; 1284 days = iso_week_days (tp->tm_yday + (365 + __isleap (year)), 1285 tp->tm_wday); 1286 } 1287 else 1288 { 1289 int d = iso_week_days (tp->tm_yday - (365 + __isleap (year)), 1290 tp->tm_wday); 1291 if (0 <= d) 1292 { 1293 /* This ISO week belongs to the next year. */ 1294 year++; 1295 days = d; 1296 } 1297 } 1298 1299 switch (*f) 1300 { 1301 case L_('g'): 1302 DO_NUMBER (2, (year % 100 + 100) % 100); 1303 1304 case L_('G'): 1305 DO_NUMBER (1, year); 1306 1307 default: 1308 DO_NUMBER (2, days / 7 + 1); 1309 } 1310 } 1311 1312 case L_('W'): 1313 if (modifier == L_('E')) 1314 goto bad_format; 1315 1316 DO_NUMBER (2, (tp->tm_yday - (tp->tm_wday - 1 + 7) % 7 + 7) / 7); 1317 1318 case L_('w'): 1319 if (modifier == L_('E')) 1320 goto bad_format; 1321 1322 DO_NUMBER (1, tp->tm_wday); 1323 1324 case L_('Y'): 1325 if (modifier == 'E') 1326 { 1327 #if HAVE_STRUCT_ERA_ENTRY 1328 struct era_entry *era = _nl_get_era_entry (tp HELPER_LOCALE_ARG); 1329 if (era) 1330 { 1331 # ifdef COMPILE_WIDE 1332 subfmt = era->era_wformat; 1333 # else 1334 subfmt = era->era_format; 1335 # endif 1336 goto subformat; 1337 } 1338 #else 1339 # if HAVE_STRFTIME 1340 goto underlying_strftime; 1341 # endif 1342 #endif 1343 } 1344 if (modifier == L_('O')) 1345 goto bad_format; 1346 else 1347 DO_NUMBER (1, tp->tm_year + TM_YEAR_BASE); 1348 1349 case L_('y'): 1350 if (modifier == L_('E')) 1351 { 1352 #if HAVE_STRUCT_ERA_ENTRY 1353 struct era_entry *era = _nl_get_era_entry (tp HELPER_LOCALE_ARG); 1354 if (era) 1355 { 1356 int delta = tp->tm_year - era->start_date[0]; 1357 DO_NUMBER (1, (era->offset 1358 + delta * era->absolute_direction)); 1359 } 1360 #else 1361 # if HAVE_STRFTIME 1362 goto underlying_strftime; 1363 # endif 1364 #endif 1365 } 1366 DO_NUMBER (2, (tp->tm_year % 100 + 100) % 100); 1367 1368 case L_('Z'): 1369 if (change_case) 1370 { 1371 to_uppcase = 0; 1372 to_lowcase = 1; 1373 } 1374 1375 #if HAVE_TZNAME 1376 /* The tzset() call might have changed the value. */ 1377 if (!(zone && *zone) && tp->tm_isdst >= 0) 1378 zone = tzname[tp->tm_isdst]; 1379 #endif 1380 if (! zone) 1381 zone = ""; 1382 1383 #ifdef COMPILE_WIDE 1384 { 1385 /* The zone string is always given in multibyte form. We have 1386 to transform it first. */ 1387 wchar_t *wczone; 1388 size_t len; 1389 widen (zone, wczone, len); 1390 cpy (len, wczone); 1391 } 1392 #else 1393 cpy (strlen (zone), zone); 1394 #endif 1395 break; 1396 1397 case L_('z'): 1398 if (tp->tm_isdst < 0) 1399 break; 1400 1401 { 1402 int diff; 1403 #if HAVE_TM_GMTOFF 1404 diff = tp->tm_gmtoff; 1405 #else 1406 if (ut) 1407 diff = 0; 1408 else 1409 { 1410 struct tm gtm; 1411 struct tm ltm; 1412 time_t lt; 1413 1414 ltm = *tp; 1415 lt = mktime (<m); 1416 1417 if (lt == (time_t) -1) 1418 { 1419 /* mktime returns -1 for errors, but -1 is also a 1420 valid time_t value. Check whether an error really 1421 occurred. */ 1422 struct tm tm; 1423 1424 if (! my_strftime_localtime_r (<, &tm) 1425 || ((ltm.tm_sec ^ tm.tm_sec) 1426 | (ltm.tm_min ^ tm.tm_min) 1427 | (ltm.tm_hour ^ tm.tm_hour) 1428 | (ltm.tm_mday ^ tm.tm_mday) 1429 | (ltm.tm_mon ^ tm.tm_mon) 1430 | (ltm.tm_year ^ tm.tm_year))) 1431 break; 1432 } 1433 1434 if (! my_strftime_gmtime_r (<, >m)) 1435 break; 1436 1437 diff = tm_diff (<m, >m); 1438 } 1439 #endif 1440 1441 if (diff < 0) 1442 { 1443 add (1, *p = L_('-')); 1444 diff = -diff; 1445 } 1446 else 1447 add (1, *p = L_('+')); 1448 1449 diff /= 60; 1450 DO_NUMBER (4, (diff / 60) * 100 + diff % 60); 1451 } 1452 1453 case L_('\0'):/* GNU extension: % at end of format. */ 1454 --f; 1455 /* Fall through. */ At conditional (15): "default" taking true path 1456 default: 1457 /* Unknown format; output the format, including the '%', 1458 since this is most likely the right thing to do if a 1459 multibyte string has been misparsed. */ 1460 bad_format: 1461 { 1462 int flen; At conditional (16): "*(f + (1 - flen)) != 37" taking true path At conditional (17): "*(f + (1 - flen)) != 37" taking true path At conditional (18): "*(f + (1 - flen)) != 37" taking true path At conditional (19): "*(f + (1 - flen)) != 37" taking false path 1463 for (flen = 1; f[1 - flen] != L_('%'); flen++) 1464 continue; At conditional (20): "_delta > 0" taking true path At conditional (21): "_incr >= (maxsize - i)" taking true path 1465 cpy (flen, &f[1 - flen]); 1466 } 1467 break; 1468 } 1469 } 1470

Re: problem report #104

Andreas Schwab — 2008-12-01T18:18:56

In practice, this never happens. Andreas.