summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
...
* Misc autoconf updatesv1.1.31-rc1Nick Wellnhofer2017-09-144-239/+15
| | | | | | | | | | | | * Switch to autoreconf in autogen.sh * Rename configure.in to configure.ac * Use modern form of AC_INIT and AM_INIT_AUTOMAKE * Set "foreign" automake option to avoid creation of a default COPYING and INSTALL. * Remove AC_PATH_PROG(RM), otherwise libtool's $RM will be overwritten and cause the infamous "cannot remove libtoolT" warning. * Remove autogenerated config.h.in from version control * Update .gitignore
* Fix pkg-config related regressions in configure scriptNick Wellnhofer2017-09-141-32/+22
| | | | | | | | | | Fix bashism in configure.in. Resulted in pkg-config never getting used for libxml if sh is not bash. Move pkg-config block to correct location. Rework libxml2 detection logic. Fixes bugs 787602 and 787617.
* Release of libxslt-1.1.30v1.1.30Daniel Veillard2017-09-0430-187/+715
| | | | | * configure.in, doc/xslt.html: updated for the release * doc/*: regenerated
* Also fix memory hazards in exsltFuncResultElemv1.1.30-rc2Nick Wellnhofer2017-08-293-5/+44
| | | | | | | | | | Similar to the previous fix to exsltFuncFunctionFunction, exsltFuncResultElem also has to get the current node from the transform context instead of the XPath context. Thanks to Nicolas Gregoire for the report. Fixes bug 786989.
* Fix NULL deref in xsltDefaultSortFunctionv1.1.30-rc1Nick Wellnhofer2017-07-315-0/+24
| | | | | | | | | An evaluation error in a secondary sort key could lead to a NULL pointer dereference. Thanks to Nicolas Gregoire for the report. Fixes bug 785588.
* Fix memory hazards in exsltFuncFunctionFunctionNick Wellnhofer2017-07-314-1/+27
| | | | | | | | | | Get the current node from the transform context. The current node in the XPath context isn't guaranteed to be preserved when evaluating the function template and could point to an invalid address. Thanks to Nicolas Gregoire for the report. Fixes bug 785589.
* Add .travis.ymlNick Wellnhofer2017-06-011-0/+32
| | | | | | | | | | | | | | | | | | For now this is mainly useful if you work on a fork of the libxslt mirror on GitHub: https://github.com/GNOME/libxslt Start with two build setups: - GCC with as many GNU extensions disabled as possible, trying to emulate a C89 compiler on a POSIX system. - clang with ASan and UBSan. The test suite doesn't set an exit code on failure, so log the test output and grep for unexpected lines. This doesn't work for the Python tests yet, so they're still disabled.
* Silence tests a littleNick Wellnhofer2017-06-014-32/+22
|
* Set LIBXML_SRC to absolute pathNick Wellnhofer2017-06-011-4/+4
| | | | Fixes Python tests when using --with-libxml-src with a relative path.
* Add missing #includeNick Wellnhofer2017-06-011-0/+1
|
* Adjust expected error messages in testsNick Wellnhofer2017-05-275-7/+1
| | | | libxml2 commit c851970 removed some useless error messages.
* Fix memory leaks in EXSLT error pathsNick Wellnhofer2017-05-273-36/+19
| | | | Found with libFuzzer and ASan.
* Fix memory leak in str:concat with empty node-setNick Wellnhofer2017-05-271-0/+1
| | | | Found with libFuzzer and ASan.
* Fix memory leaks in error pathsNick Wellnhofer2017-05-274-2/+10
| | | | Found with libFuzzer and ASan.
* Make xsltDebug more quietNick Wellnhofer2017-05-271-1/+4
| | | | | Only call xmlXPathDebugDumpObject if the debug context is stdout or stderr.
* Switch to xmlUTF8Strsize in numbers.cNick Wellnhofer2017-05-271-45/+24
| | | | | | | | | When encountering invalid UTF-8, xsltUTF8Size can return a size greater than the actual string length or -1. Switch to xmlUTF8Strsize which returns a sensible size even with invalid UTF-8. Under normal conditions, libxslt should never receive invalid UTF-8. But this change helps when fuzzing and hardens security.
* Fix NULL pointer deref in xsltFormatNumberFunctionNick Wellnhofer2017-05-271-1/+3
| | | | | | | | Introduced recently with my commit 38d4a90 Make xsl:decimal-format work with namespaces Coverity CID 1434477. Also found with libFuzzer and ASan.
* Fix UTF-8 check in str:paddingNick Wellnhofer2017-05-271-4/+5
| | | | | Make sure that all arguments are popped before checking for UTF-8 validity. Improves upon recent commit 1785d11.
* Look for libxml2 via pkg-config firstElliott Sales de Andrade2017-05-181-1/+25
| | | | | | | | | | | | | | | | | | | | | If no explicit path is specified, try pkg-config first, before xml2-config. The reason is that pkg-config knows the difference between static and shared dependencies and thus doesn't cause libxslt to be linked against a bunch of extra stuff. Say for example that libxml2 is compiled --with-icu, then it will be linked against various libicu shared libraries. xml2-config will _also_ specify those libraries (because it doesn't know whether you are doing static or shared linking) and thus libxslt is also linked against libicu even though it does not use it. On the other hand, pkg-config has Libs/Libs.private which separates shared&static dependencies and so you can get libxslt to link to _only_ libxml2 without any other things. Fixes bug #778549: https://bugzilla.gnome.org/show_bug.cgi?id=778549
* New-line terminate error message that missed this conventionJan Pokorný2017-05-181-1/+1
| | | | | | | That could confuse library users that set their own error handler, because there are already cases in libxslt that push a single message in chunks (the same way as libxml2 does) and the user may be interested in performing the message reconstruction on its own.
* Fix xmlStrPrintf argumentNick Wellnhofer2017-05-181-2/+1
| | | | | | | | The type of the msg parameter of xmlStrPrintf has changed. See the following libxml2 commit: 4472c3a Fix some format string warnings with possible format string vulnerability
* Use xmlBuffers in EXSLT string functionsNick Wellnhofer2017-05-181-7/+27
| | | | | This improves performance with (pathologically) long strings. Make sure to use a fast allocation scheme.
* Switch to xmlUTF8Strsize in EXSLT string functionsNick Wellnhofer2017-05-181-3/+3
| | | | | | | | | When encountering invalid UTF-8, xmlUTF8Size can return a size greater than the actual string length or -1. Switch to xmlUTF8Strsize which returns a sensible size even with invalid UTF-8. Under normal conditions, libxslt should never receive invalid UTF-8. But this change helps when fuzzing and hardens security.
* Check for return value of xmlUTF8StrlenNick Wellnhofer2017-05-181-2/+25
| | | | | | | Check whether xmlUTF8Strlen returns -1 for invalid UTF-8. Under normal conditions, libxslt should never receive invalid UTF-8. But this change helps when fuzzing and hardens security.
* Avoid double/long round trip in FORMAT_ITEMNick Wellnhofer2017-05-181-8/+6
|
* Check for overflow in _exsltDateParseGYearNick Wellnhofer2017-05-181-0/+2
| | | | | | Fixes bug #765379: https://bugzilla.gnome.org/show_bug.cgi?id=765379
* Fix double to int conversionNick Wellnhofer2017-05-182-21/+35
| | | | | | Add range checks to avoid undefined behavior. Limit str:padding length to 100,000 chars.
* Separate date and duration structsNick Wellnhofer2017-05-181-284/+293
|
* Check for overflow in exsltDateParseDurationNick Wellnhofer2017-05-185-97/+102
| | | | | | | | Also fix parsing of duSecondFrag, see https://www.w3.org/TR/xmlschema11-2/#nt-duSeFrag Fix memory leak in error case.
* Check for overflow in _exsltDateDifferenceNick Wellnhofer2017-05-181-14/+29
|
* Clamp seconds field of durationsNick Wellnhofer2017-05-181-52/+58
| | | | | | | Clamp seconds field of durations to range 0..SECS_PER_DAY, simplifying some calculations. Also add overflow checks in _exsltDateAddDurCalc.
* Change _exsltDateAddDurCalc parameter typesNick Wellnhofer2017-05-181-15/+14
|
* Fix date:difference with time zonesNick Wellnhofer2017-05-183-4/+16
| | | | | With time zones, the time difference can be more than two days, requiring a modulo operation.
* Rework division/remainder arithmetic in date.cNick Wellnhofer2017-05-181-73/+96
| | | | | Use integer arithmetic where possible. Remove MODULO and FQUOTIENT macros. Simplify _exsltDateAddDurCalc.
* Remove exsltDateCastDateToNumberNick Wellnhofer2017-05-181-37/+6
| | | | This function was only used with durations.
* Change internal representation of yearsNick Wellnhofer2017-05-1813-61/+78
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | XML Schema Part 2 doesn't allow the year 0000 which seems to imply that year 0001 is preceded by -0001. The old code followed this convention but it represented the year -0001 as -1, requiring some adjustments when crossing the beginning of year 0001. Now the year -0001 is represented by 0 internally (astronomical year numbering). This simplifies some calculations. As a side effect, (XML Schema) years -0001, -0005, ... are now leap years. Previously, years -0004, -0008, ... were leap years. The new behavior seems more correct and better matches other implementations of the proleptic Gregorian calendar. Also fixes some bugs: - Previously, date:day-in-week() returned wrong values for dates before the year 3 BC. For example, it returned 6 (Friday) for both '-0004-12-31' and '-0003-01-01'. Now it returns 4 (Wednesday) for '-0004-12-31' and 5 (Thursday) for '-0003-01-01' (because of the leap year change). - date:add could return wrong results when crossing AD 1. For example, date:add('-0001-01-01', 'P2Y') would return '0001-01-01' instead of '0002-01-01'. - Likewise, date:difference produced wrong results when working on years or yearMonths.
* Optimize IS_LEAPNick Wellnhofer2017-05-171-1/+1
| | | | See http://stackoverflow.com/a/11595914
* Change version of xsltMaxVars back to 1.0.24Nick Wellnhofer2017-03-152-4/+4
| | | | | | | | | | xsltMaxVars was added in 1.1.27 with a wrong version. We have to keep the wrong version to not break the ABI. Also regenerate libxslt.syms. Fixes bug #780089: https://bugzilla.gnome.org/show_bug.cgi?id=780089
* Link libraries with libmJussi Kukkonen2017-02-102-2/+2
| | | | | | | | | Otherwise linking the resulting libraries to a binary (e.g. xsltproc) fails when using gold linker: | ../libxslt/.libs/libxslt.so: error: undefined reference to 'fmod' | ../libxslt/.libs/libxslt.so: error: undefined reference to 'pow' | ../libexslt/.libs/libexslt.so: error: undefined reference to 'floor' | collect2: error: ld returned 1 exit status
* Disable xsltCopyTextString optimization for extensionsNick Wellnhofer2017-02-101-0/+24
| | | | | | | | | | Extensions could append text using xmlAddChild which will free the buffer pointed to by 'lasttext'. This buffer could later be reallocated with a different size than recorded in 'lasttsize'. Fixes bug #777432: https://bugzilla.gnome.org/show_bug.cgi?id=777432
* Create DOCTYPE for HTML version 5Nick Wellnhofer2017-02-074-0/+13
| | | | | | | Make stylesheets with <xsl:output method="html" version="5"/> generate a HTML5 doctype. Fixes bug #778192: https://bugzilla.gnome.org/show_bug.cgi?id=778192
* Make xsl:decimal-format work with namespacesNick Wellnhofer2017-02-0719-10/+122
| | | | | | Fixes bug #778170: https://bugzilla.gnome.org/show_bug.cgi?id=778170
* Remove norm:localTime extension functionNick Wellnhofer2017-02-034-151/+1
| | | | | | | | | | | | | | | | | | | | The length of the input string wasn't checked, resulting in a minor information leak. This extension function was non-standard and undocumented and used a custom date format, so it seems best to remove it. Note that with the fix to bug #758192, it's possible to convert between time zones using date:add and date:difference: date:add($tz, date:difference($tz, $date-time)) $tz is an arbitrary dateTime in the target time zone. If you want to convert to the local time zone, set $tz to the current dateTime returned by date:date-time(). Originally reported to Chromium: https://crbug.com/670720
* Check for integer overflow in xsltAddTextStringNick Wellnhofer2017-02-032-5/+24
| | | | | | | | | Limit buffer size in xsltAddTextString to INT_MAX. The issue can be exploited to trigger an out of bounds write on 64-bit systems. Originally reported to Chromium: https://crbug.com/676623
* Detect infinite recursion when evaluating function argumentsNick Wellnhofer2017-01-165-2/+43
| | | | | | | | | | | | This fixes a regression introduced when consolidating recursion checks in commit 1c8e0e5. When a function is called recursively during evaluation of its arguments, the recursion check in xsltApplySequenceConstructor is never reached. Readd recursion check in exsltFuncFunctionFunction but use the template depth counter. Fixes bug #777293: https://bugzilla.gnome.org/show_bug.cgi?id=777293
* Rename xsltCopyTreeInternal to xsltCopyTreeNick Wellnhofer2017-01-031-45/+15
| | | | The old xsltCopyTree was unused.
* Fix memory leak in xsltElementAvailableFunctionNick Wellnhofer2017-01-031-1/+1
|
* Fix for pattern predicates calling functionsNick Wellnhofer2017-01-036-13/+39
| | | | | | | | | | Set correct XSLT instruction when evaluating predicates in patterns. This is needed by functions like element-available. Could also lead to a NULL pointer dereference. Thanks to Wang Junjie for the report. Fixes bug #776792: https://bugzilla.gnome.org/show_bug.cgi?id=776792
* Fix cmd.exe invocations in Makefile.mingwNick Wellnhofer2017-01-031-29/+29
| | | | | | | | The /C parameter should be quoted. Fixes bug #774089: https://bugzilla.gnome.org/show_bug.cgi?id=774089
* Change default SOPREFIX on Windows to "bin"Nick Wellnhofer2017-01-031-1/+1
| | | | | | | | Under MinGW, the convention is to install DLLs under bin. Fixes bug #774090: https://bugzilla.gnome.org/show_bug.cgi?id=774090
View Lorry Controller Status