From 424e4173ba761870934bb350303e383fffc0e01c Mon Sep 17 00:00:00 2001 From: Jean-Francois Dockes Date: Wed, 28 Sep 2011 15:01:14 +0200 Subject: [PATCH] threading cleanup: add mutex protection around moronic change to transcode. Add mutex to equiv issue in unac. Rename const strings everywhere to cstr_xx to ease future detection of potentially problematic static variables. Most probably close issue #65 --- src/aspell/rclaspell.cpp | 6 +- src/aspell/rclaspell.h | 2 +- src/common/rclinit.cpp | 3 + src/index/beaglequeue.cpp | 2 - src/internfile/internfile.cpp | 75 ++++++++++++------------ src/internfile/mh_mbox.cpp | 36 ++++++++---- src/query/reslistpager.cpp | 18 +++--- src/rcldb/rcldb.cpp | 69 ++++++++++++---------- src/rcldb/rclquery.cpp | 9 +-- src/rcldb/searchdata.cpp | 6 +- src/rcldb/stemdb.cpp | 8 +-- src/unac/unac.c | 44 ++++++++++---- src/unac/unac.h | 3 + src/utils/ptmutex.cpp | 106 ++++++++++++++++++++++++++++++++++ src/utils/pxattr.cpp | 16 ++--- src/utils/smallut.cpp | 7 ++- src/utils/transcode.cpp | 21 +++++-- unac/unac.c | 44 ++++++++++---- unac/unac.h | 3 + website/BUGS.html | 14 ++++- website/download.html | 46 ++++++++++++--- website/index.html.en | 24 ++++++++ 22 files changed, 411 insertions(+), 151 deletions(-) create mode 100644 src/utils/ptmutex.cpp diff --git a/src/aspell/rclaspell.cpp b/src/aspell/rclaspell.cpp index 0874118d..d08b6ad1 100644 --- a/src/aspell/rclaspell.cpp +++ b/src/aspell/rclaspell.cpp @@ -34,6 +34,8 @@ #include "rclaspell.h" #include "debuglog.h" +#include "ptmutex.h" + // Just a place where we keep the Aspell library entry points together class AspellApi { public: @@ -61,6 +63,7 @@ public: }; static AspellApi aapi; +static PTMutexInit o_aapi_mutex; #define NMTOPTR(NM, TP) \ if ((aapi.NM = TP dlsym(m_data->m_handle, #NM)) == 0) { \ @@ -111,6 +114,7 @@ Aspell::~Aspell() bool Aspell::init(string &reason) { + PTMutexLocker locker(o_aapi_mutex); deleteZ(m_data); // Language: we get this from the configuration, else from the NLS @@ -227,7 +231,7 @@ bool Aspell::init(string &reason) return true; } -bool Aspell::ok() +bool Aspell::ok() const { return m_data != 0 && m_data->m_handle != 0; } diff --git a/src/aspell/rclaspell.h b/src/aspell/rclaspell.h index 9f2db884..06032495 100644 --- a/src/aspell/rclaspell.h +++ b/src/aspell/rclaspell.h @@ -50,7 +50,7 @@ class Aspell { ~Aspell(); /** Check health */ - bool ok(); + bool ok() const; /** Find the aspell command and shared library, init function pointers */ bool init(string &reason); diff --git a/src/common/rclinit.cpp b/src/common/rclinit.cpp index c2ec8728..fdfa7747 100644 --- a/src/common/rclinit.cpp +++ b/src/common/rclinit.cpp @@ -29,6 +29,7 @@ #include "rclconfig.h" #include "rclinit.h" #include "pathut.h" +#include "unac.h" static const int catchedSigs[] = {SIGHUP, SIGINT, SIGQUIT, SIGTERM, SIGUSR1, SIGUSR2}; @@ -99,6 +100,8 @@ RclConfig *recollinit(RclInitFlags flags, // Make sure the locale charset is initialized (so that multiple // threads don't try to do it at once). config->getDefCharset(); + // Init unac locking + unac_init_mt(); int flushmb; if (config->getConfParam("idxflushmb", &flushmb) && flushmb > 0) { diff --git a/src/index/beaglequeue.cpp b/src/index/beaglequeue.cpp index b7e19e2d..56f6d0dc 100644 --- a/src/index/beaglequeue.cpp +++ b/src/index/beaglequeue.cpp @@ -172,8 +172,6 @@ public: ifstream m_input; }; -const string badtmpdirname = "/no/such/dir/really/can/exist"; - // Initialize. Compute paths and create a temporary directory that will be // used by internfile() BeagleQueueIndexer::BeagleQueueIndexer(RclConfig *cnf, Rcl::Db *db, diff --git a/src/internfile/internfile.cpp b/src/internfile/internfile.cpp index 516eb003..2014c9b9 100644 --- a/src/internfile/internfile.cpp +++ b/src/internfile/internfile.cpp @@ -51,20 +51,21 @@ using namespace std; #include "pxattr.h" #endif // RCL_USE_XATTR -static const string stxtplain("text/plain"); +static const string cstr_stxtplain("text/plain"); // The internal path element separator. This can't be the same as the rcldb // file to ipath separator : "|" // We replace it with a control char if it comes out of a filter (ie: // rclzip or rclchm can do this). If you want the SOH control char // inside an ipath, you're out of luck (and a bit weird). -static const string isep(":"); -static const char colon_repl = '\x01'; +static const string cstr_isep(":"); + +static const char cchar_colon_repl = '\x01'; static string colon_hide(const string& in) { string out; for (string::const_iterator it = in.begin(); it != in.end(); it++) { - out += *it == ':' ? colon_repl : *it; + out += *it == ':' ? cchar_colon_repl : *it; } return out; } @@ -72,7 +73,7 @@ static string colon_restore(const string& in) { string out; for (string::const_iterator it = in.begin(); it != in.end(); it++) { - out += *it == colon_repl ? ':' : *it; + out += *it == cchar_colon_repl ? ':' : *it; } return out; } @@ -115,7 +116,7 @@ bool FileInterner::getEnclosing(const string &url, const string &ipath, url.c_str(), eipath.c_str())); if (eipath.empty()) return false; - if ((colon = eipath.find_last_of(isep)) != string::npos) { + if ((colon = eipath.find_last_of(cstr_isep)) != string::npos) { eipath.erase(colon); } else { eipath.erase(); @@ -365,12 +366,12 @@ void FileInterner::initcommon(RclConfig *cnf, int flags) m_handlers.reserve(MAXHANDLERS); for (unsigned int i = 0; i < MAXHANDLERS; i++) m_tmpflgs[i] = false; - m_targetMType = stxtplain; + m_targetMType = cstr_stxtplain; } // We used a single beagle cache object to access beagle data. We protect it // against multiple thread access. -static PTMutexInit o_lock; +static PTMutexInit o_beagler_mutex; FileInterner::FileInterner(const Rcl::Doc& idoc, RclConfig *cnf, TempDir& td, int flags) @@ -422,12 +423,12 @@ FileInterner::FileInterner(const Rcl::Doc& idoc, RclConfig *cnf, string udi = it->second; { - PTMutexLocker locker(o_lock); + PTMutexLocker locker(o_beagler_mutex); // Retrieve from our webcache (beagle data). The beagler // object is created at the first call of this routine and // deleted when the program exits. - static BeagleQueueCache beagler(cnf); - if (!beagler.getFromCache(udi, dotdoc, data)) { + static BeagleQueueCache o_beagler(cnf); + if (!o_beagler.getFromCache(udi, dotdoc, data)) { LOGINFO(("FileInterner:: failed fetch from Beagle cache for [%s]\n", udi.c_str())); return; @@ -564,14 +565,14 @@ static inline bool getKeyValue(const map& docdata, // These defs are for the Dijon meta array. Rcl::Doc predefined field // names are used where appropriate. In some cases, Rcl::Doc names are // used inside the Dijon metadata (ex: origcharset) -static const string keyau("author"); -static const string keycs("charset"); -static const string keyct("content"); -static const string keyds("description"); -static const string keyfn("filename"); -static const string keymd("modificationdate"); -static const string keymt("mimetype"); -static const string keytt("title"); +static const string cstr_keyau("author"); +static const string cstr_keycs("charset"); +static const string cstr_keyct("content"); +static const string cstr_keyds("description"); +static const string cstr_keyfn("filename"); +static const string cstr_keymd("modificationdate"); +static const string cstr_keymt("mimetype"); +static const string cstr_keytt("title"); bool FileInterner::dijontorcl(Rcl::Doc& doc) { @@ -585,21 +586,21 @@ bool FileInterner::dijontorcl(Rcl::Doc& doc) for (map::const_iterator it = docdata.begin(); it != docdata.end(); it++) { - if (it->first == keyct) { + if (it->first == cstr_keyct) { doc.text = it->second; - } else if (it->first == keymd) { + } else if (it->first == cstr_keymd) { doc.dmtime = it->second; } else if (it->first == Rcl::Doc::keyoc) { doc.origcharset = it->second; - } else if (it->first == keymt || it->first == keycs) { + } else if (it->first == cstr_keymt || it->first == cstr_keycs) { // don't need/want these. } else { doc.meta[it->first] = it->second; } } - if (doc.meta[Rcl::Doc::keyabs].empty() && !doc.meta[keyds].empty()) { - doc.meta[Rcl::Doc::keyabs] = doc.meta[keyds]; - doc.meta.erase(keyds); + if (doc.meta[Rcl::Doc::keyabs].empty() && !doc.meta[cstr_keyds].empty()) { + doc.meta[Rcl::Doc::keyabs] = doc.meta[cstr_keyds]; + doc.meta.erase(cstr_keyds); } return true; } @@ -635,21 +636,21 @@ void FileInterner::collectIpathAndMT(Rcl::Doc& doc) const if (!ipathel.empty()) { // We have a non-empty ipath hasipath = true; - getKeyValue(docdata, keymt, doc.mimetype); - getKeyValue(docdata, keyfn, doc.utf8fn); + getKeyValue(docdata, cstr_keymt, doc.mimetype); + getKeyValue(docdata, cstr_keyfn, doc.utf8fn); } - doc.ipath += colon_hide(ipathel) + isep; + doc.ipath += colon_hide(ipathel) + cstr_isep; } else { - doc.ipath += isep; + doc.ipath += cstr_isep; } - getKeyValue(docdata, keyau, doc.meta[Rcl::Doc::keyau]); - getKeyValue(docdata, keymd, doc.dmtime); + getKeyValue(docdata, cstr_keyau, doc.meta[Rcl::Doc::keyau]); + getKeyValue(docdata, cstr_keymd, doc.dmtime); } // Trim empty tail elements in ipath. if (hasipath) { LOGDEB2(("IPATH [%s]\n", doc.ipath.c_str())); - string::size_type sit = doc.ipath.find_last_not_of(isep); + string::size_type sit = doc.ipath.find_last_not_of(cstr_isep); if (sit == string::npos) doc.ipath.erase(); else if (sit < doc.ipath.length() -1) @@ -681,8 +682,8 @@ int FileInterner::addHandler() { const map& docdata = m_handlers.back()->get_meta_data(); string charset, mimetype; - getKeyValue(docdata, keycs, charset); - getKeyValue(docdata, keymt, mimetype); + getKeyValue(docdata, cstr_keycs, charset); + getKeyValue(docdata, cstr_keymt, mimetype); LOGDEB(("FileInterner::addHandler: next_doc is %s\n", mimetype.c_str())); @@ -690,7 +691,7 @@ int FileInterner::addHandler() // general), we're done decoding. If we hit text/plain, we're done // in any case if (!stringicmp(mimetype, m_targetMType) || - !stringicmp(mimetype, stxtplain)) { + !stringicmp(mimetype, cstr_stxtplain)) { m_reachedMType = mimetype; LOGDEB1(("FileInterner::addHandler: target reached\n")); return ADD_BREAK; @@ -723,7 +724,7 @@ int FileInterner::addHandler() const string *txt = &ns; { map::const_iterator it; - it = docdata.find(keyct); + it = docdata.find(cstr_keyct); if (it != docdata.end()) txt = &it->second; } @@ -795,7 +796,7 @@ FileInterner::Status FileInterner::internfile(Rcl::Doc& doc, const string& ipath int vipathidx = 0; if (!ipath.empty()) { vector lipath; - stringToTokens(ipath, lipath, isep, true); + stringToTokens(ipath, lipath, cstr_isep, true); for (vector::iterator it = lipath.begin(); it != lipath.end(); it++) { *it = colon_restore(*it); diff --git a/src/internfile/mh_mbox.cpp b/src/internfile/mh_mbox.cpp index b3d21ce1..5d08b1b8 100644 --- a/src/internfile/mh_mbox.cpp +++ b/src/internfile/mh_mbox.cpp @@ -53,7 +53,7 @@ public: private: FILE **m_fpp; }; -static PTMutexInit o_mutex; +static PTMutexInit o_mcache_mutex; /** * Handles a cache for message numbers to offset translations. Permits direct @@ -86,7 +86,7 @@ public: LOGDEB0(("MboxCache::get_offsets: init failed\n")); return -1; } - PTMutexLocker locker(o_mutex); + PTMutexLocker locker(o_mcache_mutex); string fn = makefilename(udi); FILE *fp = 0; if ((fp = fopen(fn.c_str(), "r")) == 0) { @@ -133,7 +133,7 @@ public: return; if (fsize < m_minfsize) return; - PTMutexLocker locker(o_mutex); + PTMutexLocker locker(o_mcache_mutex); string fn = makefilename(udi); FILE *fp; if ((fp = fopen(fn.c_str(), "w")) == 0) { @@ -163,7 +163,7 @@ public: // Check state, possibly initialize bool ok(RclConfig *config) { - PTMutexLocker locker(o_mutex); + PTMutexLocker locker(o_mcache_mutex); if (m_minfsize == -1) return false; if (!m_ok) { @@ -224,9 +224,9 @@ private: const size_t MboxCache::o_b1size = 1024; -static class MboxCache mcache; +static class MboxCache o_mcache; -static const string keyquirks("mhmboxquirks"); +static const string cstr_keyquirks("mhmboxquirks"); MimeHandlerMbox::~MimeHandlerMbox() { @@ -271,7 +271,7 @@ bool MimeHandlerMbox::set_document_file(const string &fn) // Check for location-based quirks: string quirks; - if (m_config && m_config->getConfParam(keyquirks, quirks)) { + if (m_config && m_config->getConfParam(cstr_keyquirks, quirks)) { if (quirks == "tbird") { LOGDEB(("MimeHandlerMbox: setting quirks TBIRD\n")); m_quirks |= MBOXQUIRK_TBIRD; @@ -358,6 +358,20 @@ static const char *miniTbirdFrom = "^From $"; static regex_t fromregex; static regex_t minifromregex; static bool regcompiled; +static PTMutexInit o_regex_mutex; + +static void compileregexes() +{ + PTMutexLocker locker(o_regex_mutex); + // As the initial test of regcompiled is unprotected the value may + // have changed while we were waiting for the lock. Test again now + // that we are alone. + if (regcompiled) + return; + regcomp(&fromregex, frompat, REG_NOSUB|REG_EXTENDED); + regcomp(&minifromregex, miniTbirdFrom, REG_NOSUB|REG_EXTENDED); + regcompiled = true; +} bool MimeHandlerMbox::next_document() { @@ -383,9 +397,7 @@ bool MimeHandlerMbox::next_document() mtarg = -1; if (!regcompiled) { - regcomp(&fromregex, frompat, REG_NOSUB|REG_EXTENDED); - regcomp(&minifromregex, miniTbirdFrom, REG_NOSUB|REG_EXTENDED); - regcompiled = true; + compileregexes(); } // If we are called to retrieve a specific message, seek to bof @@ -403,7 +415,7 @@ bool MimeHandlerMbox::next_document() LOGDEB0(("MimeHandlerMbox::next_doc: mtarg %d m_udi[%s]\n", mtarg, m_udi.c_str())); if (!m_udi.empty() && - (off = mcache.get_offset(m_config, m_udi, mtarg)) >= 0 && + (off = o_mcache.get_offset(m_config, m_udi, mtarg)) >= 0 && fseeko(fp, (off_t)off, SEEK_SET) >= 0 && fgets(line, LL, fp) && (!regexec(&fromregex, line, 0, 0, 0) || @@ -492,7 +504,7 @@ bool MimeHandlerMbox::next_document() LOGDEB2(("MimeHandlerMbox::next: eof hit\n")); m_havedoc = false; if (!m_udi.empty() && storeoffsets) { - mcache.put_offsets(m_config, m_udi, m_fsize, m_offsets); + o_mcache.put_offsets(m_config, m_udi, m_fsize, m_offsets); } } return msgtxt.empty() ? false : true; diff --git a/src/query/reslistpager.cpp b/src/query/reslistpager.cpp index 1f04aa5f..f50fa074 100644 --- a/src/query/reslistpager.cpp +++ b/src/query/reslistpager.cpp @@ -34,12 +34,14 @@ using std::endl; #include "plaintorich.h" #include "mimehandler.h" -// Default highlighter +// Default highlighter. No need for locking, this is query-only. +static const string cstr_hlfontcolor(""); +static const string cstr_hlendfont(""); class PlainToRichHtReslist : public PlainToRich { public: virtual ~PlainToRichHtReslist() {} - virtual string startMatch() {return string("");} - virtual string endMatch() {return string("");} + virtual string startMatch() {return cstr_hlfontcolor;} + virtual string endMatch() {return cstr_hlendfont;} }; static PlainToRichHtReslist g_hiliter; @@ -406,10 +408,10 @@ string ResListPager::detailsLink() const string &ResListPager::parFormat() { - static const string format("" - "%R %S %L   %T
" - "%M %D   %U
" - "%A %K"); - return format; + static const string cstr_format("" + "%R %S %L   %T
" + "%M %D   %U
" + "%A %K"); + return cstr_format; } diff --git a/src/rcldb/rcldb.cpp b/src/rcldb/rcldb.cpp index 59ada1ff..96fb24e5 100644 --- a/src/rcldb/rcldb.cpp +++ b/src/rcldb/rcldb.cpp @@ -51,6 +51,7 @@ using namespace std; #include "md5.h" #include "rclversion.h" #include "cancelcheck.h" +#include "ptmutex.h" #ifndef MAX #define MAX(A,B) (A>B?A:B) @@ -61,8 +62,8 @@ using namespace std; // Recoll index format version is stored in user metadata. When this change, // we can't open the db and will have to reindex. -static const string RCL_IDX_VERSION_KEY("RCL_IDX_VERSION_KEY"); -static const string RCL_IDX_VERSION("1"); +static const string cstr_RCL_IDX_VERSION_KEY("RCL_IDX_VERSION_KEY"); +static const string cstr_RCL_IDX_VERSION("1"); // This is the word position offset at which we index the body text // (abstract, keywords, etc.. are stored before this) @@ -79,7 +80,7 @@ const string end_of_field_term = "XXND"; // This is used as a marker inside the abstract frag lists, but // normally doesn't remain in final output (which is built with a // custom sep. by our caller). -static const string ellipsis("..."); +static const string cstr_ellipsis("..."); string version_string(){ return string("Recoll ") + string(rclversionstr) + string(" + Xapian ") + @@ -88,12 +89,12 @@ string version_string(){ // Synthetic abstract marker (to discriminate from abstract actually // found in document) -static const string rclSyntAbs("?!#@"); +static const string cstr_syntAbs("?!#@"); // Only ONE field name inside the index data record differs from the // Rcl::Doc ones: caption<->title, for a remnant of compatibility with // omega -static const string keycap("caption"); +static const string cstr_keycap("caption"); // Static/Default table for field->prefix/weight translation. // This is logically const after initialization. Can't use a @@ -106,8 +107,16 @@ static const string keycap("caption"); // suppressed. static map fldToTraits; +static PTMutexInit o_fldToTraits_mutex; + static void initFldToTraits() { + PTMutexLocker locker(o_fldToTraits_mutex); + // As we perform non-locked testing of initialization, check again with + // the lock held + if (fldToTraits.size()) + return; + // Can't remember why "abstract" is indexed without a prefix // (result: it's indexed twice actually). Maybe I'll dare change // this one day @@ -116,7 +125,7 @@ static void initFldToTraits() fldToTraits["ext"] = FieldTraits("XE"); fldToTraits[Doc::keyfn] = FieldTraits("XSFN"); - fldToTraits[keycap] = FieldTraits("S"); + fldToTraits[cstr_keycap] = FieldTraits("S"); fldToTraits[Doc::keytt] = FieldTraits("S"); fldToTraits["subject"] = FieldTraits("S"); @@ -189,14 +198,14 @@ bool Db::Native::dbDataToRclDoc(Xapian::docid docid, std::string &data, parms.get(Doc::keyfmt, doc.fmtime); parms.get(Doc::keydmt, doc.dmtime); parms.get(Doc::keyoc, doc.origcharset); - parms.get(keycap, doc.meta[Doc::keytt]); + parms.get(cstr_keycap, doc.meta[Doc::keytt]); parms.get(Doc::keykw, doc.meta[Doc::keykw]); parms.get(Doc::keyabs, doc.meta[Doc::keyabs]); // Possibly remove synthetic abstract indicator (if it's there, we // used to index the beginning of the text as abstract). doc.syntabs = false; - if (doc.meta[Doc::keyabs].find(rclSyntAbs) == 0) { - doc.meta[Doc::keyabs] = doc.meta[Doc::keyabs].substr(rclSyntAbs.length()); + if (doc.meta[Doc::keyabs].find(cstr_syntAbs) == 0) { + doc.meta[Doc::keyabs] = doc.meta[Doc::keyabs].substr(cstr_syntAbs.length()); doc.syntabs = true; } parms.get(Doc::keyipt, doc.ipath); @@ -417,7 +426,7 @@ vector Db::Native::makeAbstract(Xapian::docid docid, Query *query) } else if (ii > (unsigned int)ipos && ii < (unsigned int)ipos + qtrmwrdcnt) { sparseDoc[ii] = occupiedmarker; - } else if (!sparseDoc[ii].compare(ellipsis)) { + } else if (!sparseDoc[ii].compare(cstr_ellipsis)) { // For an empty slot, the test has a side // effect of inserting an empty string which // is what we want @@ -429,7 +438,7 @@ vector Db::Native::makeAbstract(Xapian::docid docid, Query *query) // empty string here, we really want an empty slot, // use find() if (sparseDoc.find(sto+1) == sparseDoc.end()) { - sparseDoc[sto+1] = ellipsis; + sparseDoc[sto+1] = cstr_ellipsis; } // Limit to allocated occurences and total size @@ -531,7 +540,7 @@ vector Db::Native::makeAbstract(Xapian::docid docid, Query *query) if (!incjk || (incjk && !newcjk)) chunk += " "; incjk = newcjk; - if (it->second == ellipsis) { + if (it->second == cstr_ellipsis) { vabs.push_back(chunk); chunk.clear(); } else { @@ -612,8 +621,8 @@ bool Db::open(OpenMode mode, OpenError *error) // If db is empty, write the data format version at once // to avoid stupid error messages: if (m_ndb->xwdb.get_doccount() == 0) - m_ndb->xwdb.set_metadata(RCL_IDX_VERSION_KEY, - RCL_IDX_VERSION); + m_ndb->xwdb.set_metadata(cstr_RCL_IDX_VERSION_KEY, + cstr_RCL_IDX_VERSION); m_ndb->m_iswritable = true; // We open a readonly object in all cases (possibly in // addition to the r/w one) because some operations @@ -650,11 +659,11 @@ bool Db::open(OpenMode mode, OpenError *error) // Check index format version. Must not try to check a just created or // truncated db if (mode != DbTrunc && m_ndb->xdb().get_doccount() > 0) { - string version = m_ndb->xdb().get_metadata(RCL_IDX_VERSION_KEY); - if (version.compare(RCL_IDX_VERSION)) { + string version = m_ndb->xdb().get_metadata(cstr_RCL_IDX_VERSION_KEY); + if (version.compare(cstr_RCL_IDX_VERSION)) { m_ndb->m_noversionwrite = true; LOGERR(("Rcl::Db::open: file index [%s], software [%s]\n", - version.c_str(), RCL_IDX_VERSION.c_str())); + version.c_str(), cstr_RCL_IDX_VERSION.c_str())); throw Xapian::DatabaseError("Recoll index version mismatch", "", ""); } @@ -693,7 +702,7 @@ bool Db::i_close(bool final) bool w = m_ndb->m_iswritable; if (w) { if (!m_ndb->m_noversionwrite) - m_ndb->xwdb.set_metadata(RCL_IDX_VERSION_KEY, RCL_IDX_VERSION); + m_ndb->xwdb.set_metadata(cstr_RCL_IDX_VERSION_KEY, cstr_RCL_IDX_VERSION); LOGDEB(("Rcl::Db:close: xapian will close. May take some time\n")); } // Used to do a flush here. Cant see why it should be necessary. @@ -952,7 +961,7 @@ void Db::setAbstractParams(int idxtrunc, int syntlen, int syntctxlen) } static const int MB = 1024 * 1024; -static const string nc("\n\r\x0c"); +static const string cstr_nc("\n\r\x0c"); #define RECORD_APPEND(R, NM, VAL) {R += NM + "=" + VAL + "\n";} @@ -1168,13 +1177,13 @@ bool Db::addOrUpdate(const string &udi, const string &parent_udi, if (doc.meta[Doc::keytt].empty()) doc.meta[Doc::keytt] = doc.utf8fn; doc.meta[Doc::keytt] = - neutchars(truncate_to_word(doc.meta[Doc::keytt], 150), nc); + neutchars(truncate_to_word(doc.meta[Doc::keytt], 150), cstr_nc); if (!doc.meta[Doc::keytt].empty()) - RECORD_APPEND(record, keycap, doc.meta[Doc::keytt]); + RECORD_APPEND(record, cstr_keycap, doc.meta[Doc::keytt]); trimstring(doc.meta[Doc::keykw], " \t\r\n"); doc.meta[Doc::keykw] = - neutchars(truncate_to_word(doc.meta[Doc::keykw], 300), nc); + neutchars(truncate_to_word(doc.meta[Doc::keykw], 300), cstr_nc); if (!doc.meta[Doc::keykw].empty()) RECORD_APPEND(record, Doc::keykw, doc.meta[Doc::keykw]); @@ -1189,12 +1198,12 @@ bool Db::addOrUpdate(const string &udi, const string &parent_udi, if (doc.meta[Doc::keyabs].empty()) { syntabs = true; if (!doc.text.empty()) - doc.meta[Doc::keyabs] = rclSyntAbs + - neutchars(truncate_to_word(doc.text, m_idxAbsTruncLen), nc); + doc.meta[Doc::keyabs] = cstr_syntAbs + + neutchars(truncate_to_word(doc.text, m_idxAbsTruncLen), cstr_nc); } else { doc.meta[Doc::keyabs] = neutchars(truncate_to_word(doc.meta[Doc::keyabs], m_idxAbsTruncLen), - nc); + cstr_nc); } if (!doc.meta[Doc::keyabs].empty()) RECORD_APPEND(record, Doc::keyabs, doc.meta[Doc::keyabs]); @@ -1205,7 +1214,7 @@ bool Db::addOrUpdate(const string &udi, const string &parent_udi, string nm = m_config->fieldCanon(*it); if (!doc.meta[*it].empty()) { string value = - neutchars(truncate_to_word(doc.meta[*it], 150), nc); + neutchars(truncate_to_word(doc.meta[*it], 150), cstr_nc); RECORD_APPEND(record, nm, value); } } @@ -1611,8 +1620,8 @@ static void addPrefix(list& terms, const string& prefix) // Characters that can begin a wildcard or regexp expression. We use skipto // to begin the allterms search with terms that begin with the portion of // the input string prior to these chars. -const string wildSpecChars = "*?["; -const string regSpecChars = "(.[{"; +const string cstr_wildSpecChars = "*?["; +const string cstr_regSpecChars = "(.[{"; // Find all index terms that match a wildcard or regular expression bool Db::termMatch(MatchType typ, const string &lang, @@ -1639,7 +1648,7 @@ bool Db::termMatch(MatchType typ, const string &lang, LOGERR(("Db::termMatch: unac failed for [%s]\n", root.c_str())); return false; } - string nochars = typ == ET_WILD ? wildSpecChars : regSpecChars; + string nochars = typ == ET_WILD ? cstr_wildSpecChars : cstr_regSpecChars; string prefix; if (!field.empty()) { @@ -1852,7 +1861,7 @@ bool Db::makeDocAbstract(Doc &doc, Query *query, string& abstract) for (vector::const_iterator it = vab.begin(); it != vab.end(); it++) { abstract.append(*it); - abstract.append(ellipsis); + abstract.append(cstr_ellipsis); } return m_reason.empty() ? true : false; } diff --git a/src/rcldb/rclquery.cpp b/src/rcldb/rclquery.cpp index d7e6e6e1..e46d7260 100644 --- a/src/rcldb/rclquery.cpp +++ b/src/rcldb/rclquery.cpp @@ -39,16 +39,17 @@ namespace Rcl { #endif +static const string cstr_keycap("caption"); +static const string cstr_keydmtime("dmtime"); + // Field names inside the index data record may differ from the rcldoc ones // (esp.: caption / title) static const string& docfToDatf(const string& df) { - static const string keycap("caption"); - static const string keydmtime("dmtime"); if (!df.compare(Doc::keytt)) { - return keycap; + return cstr_keycap; } else if (!df.compare(Doc::keymt)) { - return keydmtime; + return cstr_keydmtime; } else { return df; } diff --git a/src/rcldb/searchdata.cpp b/src/rcldb/searchdata.cpp index ad4b39d0..f7e3e0eb 100644 --- a/src/rcldb/searchdata.cpp +++ b/src/rcldb/searchdata.cpp @@ -860,13 +860,13 @@ bool StringToXapianQ::processUserString(const string &iq, return true; } -static const string nullstemlang; +static const string cstr_null; // Translate a simple OR, AND, or EXCL search clause. bool SearchDataClauseSimple::toNativeQuery(Rcl::Db &db, void *p, const string& stemlang) { - const string& l_stemlang = (m_modifiers&SDCM_NOSTEMMING)? nullstemlang: + const string& l_stemlang = (m_modifiers&SDCM_NOSTEMMING)? cstr_null: stemlang; m_terms.clear(); @@ -945,7 +945,7 @@ bool SearchDataClauseFilename::toNativeQuery(Rcl::Db &db, void *p, bool SearchDataClauseDist::toNativeQuery(Rcl::Db &db, void *p, const string& stemlang) { - const string& l_stemlang = (m_modifiers&SDCM_NOSTEMMING)? nullstemlang: + const string& l_stemlang = (m_modifiers&SDCM_NOSTEMMING)? cstr_null: stemlang; LOGDEB(("SearchDataClauseDist::toNativeQuery\n")); m_terms.clear(); diff --git a/src/rcldb/stemdb.cpp b/src/rcldb/stemdb.cpp index e5275c79..4b506a04 100644 --- a/src/rcldb/stemdb.cpp +++ b/src/rcldb/stemdb.cpp @@ -38,21 +38,21 @@ namespace Rcl { namespace StemDb { -static const string stemdirstem = "stem_"; +static const string cstr_stemdirstem = "stem_"; /// Compute name of stem db for given base database and language static string stemdbname(const string& dbdir, const string& lang) { - return path_cat(dbdir, stemdirstem + lang); + return path_cat(dbdir, cstr_stemdirstem + lang); } list getLangs(const string& dbdir) { - string pattern = stemdirstem + "*"; + string pattern = cstr_stemdirstem + "*"; list dirs = path_dirglob(dbdir, pattern); for (list::iterator it = dirs.begin(); it != dirs.end(); it++) { *it = path_basename(*it); - *it = it->substr(stemdirstem.length(), string::npos); + *it = it->substr(cstr_stemdirstem.length(), string::npos); } return dirs; } diff --git a/src/unac/unac.c b/src/unac/unac.c index 69d1ee90..92c8dea7 100644 --- a/src/unac/unac.c +++ b/src/unac/unac.c @@ -36,6 +36,7 @@ #include #include #endif /* HAVE_VSNPRINTF */ +#include #include "unac.h" #include "unac_version.h" @@ -10555,8 +10556,6 @@ int unacfold_string_utf16(const char* in, size_t in_length, outp, out_lengthp, 1); } -#define MAXOUT 1024 - static int convert(const char* from, const char* to, const char* in, size_t in_length, char** outp, size_t* out_lengthp); @@ -10564,6 +10563,14 @@ static int convert(const char* from, const char* to, static const char *utf16be = "UTF-16BE"; static iconv_t u8tou16_cd = (iconv_t)-1; static iconv_t u16tou8_cd = (iconv_t)-1; +static pthread_mutex_t o_unac_mutex; +static int unac_mutex_is_init; +// Call this or take your chances with the auto init. +void unac_init_mt() +{ + pthread_mutex_init(&o_unac_mutex, 0); + unac_mutex_is_init = 1; +} /* * Convert buffer containing string encoded in charset into @@ -10576,6 +10583,7 @@ static int convert(const char* from, const char* to, const char* in, size_t in_length, char** outp, size_t* out_lengthp) { + int ret = -1; iconv_t cd; char* out; size_t out_remain; @@ -10584,6 +10592,15 @@ static int convert(const char* from, const char* to, int from_utf16, from_utf8, to_utf16, to_utf8, u8tou16, u16tou8; const char space[] = { 0x00, 0x20 }; + /* Note: better call explicit unac_init_mt() before starting threads than + rely on this. + */ + if (unac_mutex_is_init == 0) { + pthread_mutex_init(&o_unac_mutex, 0); + unac_mutex_is_init = 1; + } + pthread_mutex_lock(&o_unac_mutex); + if (!strcmp(utf16be, from)) { from_utf8 = 0; from_utf16 = 1; @@ -10614,7 +10631,7 @@ static int convert(const char* from, const char* to, /* *outp still valid, no freeing */ if(debug_level >= UNAC_DEBUG_LOW) DEBUG("realloc %d bytes failed\n", out_size+1); - return -1; + goto out; } } else { /* +1 for null */ @@ -10622,7 +10639,7 @@ static int convert(const char* from, const char* to, if(out == 0) { if(debug_level >= UNAC_DEBUG_LOW) DEBUG("malloc %d bytes failed\n", out_size+1); - return -1; + goto out; } } out_remain = out_size; @@ -10631,7 +10648,7 @@ static int convert(const char* from, const char* to, if (u8tou16) { if (u8tou16_cd == (iconv_t)-1) { if((u8tou16_cd = iconv_open(to, from)) == (iconv_t)-1) { - return -1; + goto out; } } else { iconv(u8tou16_cd, 0, 0, 0, 0); @@ -10640,7 +10657,7 @@ static int convert(const char* from, const char* to, } else if (u16tou8) { if (u16tou8_cd == (iconv_t)-1) { if((u16tou8_cd = iconv_open(to, from)) == (iconv_t)-1) { - return -1; + goto out; } } else { iconv(u16tou8_cd, 0, 0, 0, 0); @@ -10648,7 +10665,7 @@ static int convert(const char* from, const char* to, cd = u16tou8_cd; } else { if((cd = iconv_open(to, from)) == (iconv_t)-1) { - return -1; + goto out; } } @@ -10682,7 +10699,7 @@ static int convert(const char* from, const char* to, if(errno == E2BIG) /* fall thru to the E2BIG case below */; else - return -1; + goto out; } else { /* The offending character was replaced by a SPACE, skip it. */ in += 2; @@ -10691,7 +10708,7 @@ static int convert(const char* from, const char* to, break; } } else { - return -1; + goto out; } case E2BIG: { @@ -10711,7 +10728,7 @@ static int convert(const char* from, const char* to, DEBUG("realloc %d bytes failed\n", out_size+1); free(saved); *outp = 0; - return -1; + goto out; } } out = out_base + length; @@ -10719,7 +10736,7 @@ static int convert(const char* from, const char* to, } break; default: - return -1; + goto out; break; } } @@ -10732,7 +10749,10 @@ static int convert(const char* from, const char* to, *out_lengthp = out - out_base; (*outp)[*out_lengthp] = '\0'; - return 0; + ret = 0; +out: + pthread_mutex_unlock(&o_unac_mutex); + return ret; } int unacmaybefold_string(const char* charset, diff --git a/src/unac/unac.h b/src/unac/unac.h index c447c8a1..83fb1fea 100644 --- a/src/unac/unac.h +++ b/src/unac/unac.h @@ -113,6 +113,9 @@ int unacfold_string(const char* charset, const char* in, size_t in_length, char** out, size_t* out_length); +/* To be called before starting threads in mt programs */ +void unac_init_mt(); + /* * Return unac version number. */ diff --git a/src/utils/ptmutex.cpp b/src/utils/ptmutex.cpp new file mode 100644 index 00000000..7e6e2e87 --- /dev/null +++ b/src/utils/ptmutex.cpp @@ -0,0 +1,106 @@ +/* Copyright (C) 2004 J.F.Dockes + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the + * Free Software Foundation, Inc., + * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + */ + +// +// Small test program to evaluate the cost of using mutex locks: calls +// to methods doing a small (150 bytes) base64 encoding job + string +// manips, with and without locking. The performance cost is +// negligible on all machines I tested (around 0.3% to 2% depending on +// the system and machine), but not inexistent, you would not want +// this in a tight loop. + +#include +#include +#include + +#include +using namespace std; + +#include "ptmutex.h" +#include "base64.h" + +static char *thisprog; +static char usage [] = +"ptmutex [-l] count\n" +"\n" +; +static void +Usage(void) +{ + fprintf(stderr, "%s: usage:\n%s", thisprog, usage); + exit(1); +} + +static int op_flags; +#define OPT_MOINS 0x1 +#define OPT_l 0x2 + +static const string convertbuffer = + "* The recoll GUI program sometimes crashes when running a query while\ + the indexing thread is active. Possible workarounds:"; + +static PTMutexInit o_lock; +void workerlock(string& out) +{ + PTMutexLocker locker(o_lock); + base64_encode(convertbuffer, out); +} + +void workernolock(string& out) +{ + base64_encode(convertbuffer, out); +} + +int main(int argc, char **argv) +{ + int count = 0; + thisprog = argv[0]; + argc--; argv++; + + while (argc > 0 && **argv == '-') { + (*argv)++; + if (!(**argv)) + /* Cas du "adb - core" */ + Usage(); + while (**argv) + switch (*(*argv)++) { + case 'l': op_flags |= OPT_l; break; + default: Usage(); break; + } + b1: argc--; argv++; + } + + if (argc != 1) + Usage(); + count = atoi(*argv++);argc--; + + if (op_flags & OPT_l) { + fprintf(stderr, "Looping %d, locking\n", count); + for (int i = 0; i < count; i++) { + string s; + workerlock(s); + } + } else { + fprintf(stderr, "Looping %d, no locking\n", count); + for (int i = 0; i < count; i++) { + string s; + workernolock(s); + } + } + exit(0); +} + diff --git a/src/utils/pxattr.cpp b/src/utils/pxattr.cpp index 166fa81b..83e1aab1 100644 --- a/src/utils/pxattr.cpp +++ b/src/utils/pxattr.cpp @@ -405,7 +405,7 @@ list(int fd, const string& path, vector* names, flags flags, nspace dom) return true; } -static const string nullstring(""); +static const string cstr_nullstring(""); bool get(const string& path, const string& _name, string *value, flags flags, nspace dom) @@ -414,7 +414,7 @@ bool get(const string& path, const string& _name, string *value, } bool get(int fd, const string& _name, string *value, flags flags, nspace dom) { - return get(fd, nullstring, _name, value, flags, dom); + return get(fd, cstr_nullstring, _name, value, flags, dom); } bool set(const string& path, const string& _name, const string& value, flags flags, nspace dom) @@ -424,7 +424,7 @@ bool set(const string& path, const string& _name, const string& value, bool set(int fd, const string& _name, const string& value, flags flags, nspace dom) { - return set(fd, nullstring, _name, value, flags, dom); + return set(fd, cstr_nullstring, _name, value, flags, dom); } bool del(const string& path, const string& _name, flags flags, nspace dom) { @@ -432,7 +432,7 @@ bool del(const string& path, const string& _name, flags flags, nspace dom) } bool del(int fd, const string& _name, flags flags, nspace dom) { - return del(fd, nullstring, _name, flags, dom); + return del(fd, cstr_nullstring, _name, flags, dom); } bool list(const string& path, vector* names, flags flags, nspace dom) { @@ -440,17 +440,17 @@ bool list(const string& path, vector* names, flags flags, nspace dom) } bool list(int fd, vector* names, flags flags, nspace dom) { - return list(fd, nullstring, names, flags, dom); + return list(fd, cstr_nullstring, names, flags, dom); } -static const string userstring("user."); +static const string cstr_userstring("user."); bool sysname(nspace dom, const string& pname, string* sname) { if (dom != PXATTR_USER) { errno = EINVAL; return false; } - *sname = userstring + pname; + *sname = cstr_userstring + pname; return true; } @@ -460,7 +460,7 @@ bool pxname(nspace dom, const string& sname, string* pname) errno = EINVAL; return false; } - *pname = sname.substr(userstring.length()); + *pname = sname.substr(cstr_userstring.length()); return true; } diff --git a/src/utils/smallut.cpp b/src/utils/smallut.cpp index af92128a..4b91dc04 100644 --- a/src/utils/smallut.cpp +++ b/src/utils/smallut.cpp @@ -428,7 +428,7 @@ void neutchars(const string &str, string &out, const string& chars) * if reasonably possible. Note: we could also use textsplit, stopping when * we have enough, this would be cleanly utf8-aware but would remove * punctuation */ -static const string SEPAR = " \t\n\r-:.;,/[]{}"; +static const string cstr_SEPAR = " \t\n\r-:.;,/[]{}"; string truncate_to_word(const string &input, string::size_type maxlen) { string output; @@ -436,7 +436,7 @@ string truncate_to_word(const string &input, string::size_type maxlen) output = input; } else { output = input.substr(0, maxlen); - string::size_type space = output.find_last_of(SEPAR); + string::size_type space = output.find_last_of(cstr_SEPAR); // Original version only truncated at space if space was found after // maxlen/2. But we HAVE to truncate at space, else we'd need to do // utf8 stuff to avoid truncating at multibyte char. In any case, @@ -676,6 +676,9 @@ static void gettime(int, struct m_timespec *ts) } ///// End system interface +// Note: this not protected against multithread access and not reentrant, but +// this is mostly debug code, and it won't crash, just show bad results. Also +// the frozen thing is not used that much static m_timespec frozen_tv; void Chrono::refnow() { diff --git a/src/utils/transcode.cpp b/src/utils/transcode.cpp index fcc08f4b..35ddd7fd 100644 --- a/src/utils/transcode.cpp +++ b/src/utils/transcode.cpp @@ -29,15 +29,21 @@ using std::string; #include "transcode.h" #include "debuglog.h" - +#include "ptmutex.h" #ifdef RCL_ICONV_INBUF_CONST #define ICV_P2_TYPE const char** #else #define ICV_P2_TYPE char** #endif -// We gain approximately 28% exec time for word at a time conversions by +// We gain approximately 25% exec time for word at a time conversions by // caching the iconv_open thing. +// +// We may also lose some concurrency on multiproc because of the +// necessary locking, but we only have one processing-intensive +// possible thread for now (the indexing one), so this is probably not +// an issue (and could be worked around with a slightly more +// sohisticated approach). #define ICONV_CACHE_OPEN bool transcode(const string &in, string &out, const string &icode, @@ -48,6 +54,8 @@ bool transcode(const string &in, string &out, const string &icode, static iconv_t ic = (iconv_t)-1; static string cachedicode; static string cachedocode; + static PTMutexInit o_cachediconv_mutex; + PTMutexLocker locker(o_cachediconv_mutex); #else iconv_t ic; #endif @@ -163,13 +171,14 @@ using namespace std; // Repeatedly transcode a small string for timing measurements static const string testword("\xc3\xa9\x6c\x69\x6d\x69\x6e\xc3\xa9\xc3\xa0"); -// Without cache 10e6 reps on macpro -> 1.88 S -// With cache -> 1.56 +// Without cache 10e6 reps on y -> 6.68 +// With cache -> 4.73 +// With cache and lock -> 4.9 void looptest() { cout << testword << endl; string out; - for (int i = 0; i < 1000*1000; i++) { + for (int i = 0; i < 10*1000*1000; i++) { if (!transcode(testword, out, "UTF-8", "UTF-16BE")) { cerr << "Transcode failed" << endl; break; @@ -184,7 +193,7 @@ int main(int argc, char **argv) exit(0); #endif if (argc != 5) { - cerr << "Usage: trcsguess ifilename icode ofilename ocode" << endl; + cerr << "Usage: transcode ifilename icode ofilename ocode" << endl; exit(1); } const string ifilename = argv[1]; diff --git a/unac/unac.c b/unac/unac.c index 69d1ee90..92c8dea7 100644 --- a/unac/unac.c +++ b/unac/unac.c @@ -36,6 +36,7 @@ #include #include #endif /* HAVE_VSNPRINTF */ +#include #include "unac.h" #include "unac_version.h" @@ -10555,8 +10556,6 @@ int unacfold_string_utf16(const char* in, size_t in_length, outp, out_lengthp, 1); } -#define MAXOUT 1024 - static int convert(const char* from, const char* to, const char* in, size_t in_length, char** outp, size_t* out_lengthp); @@ -10564,6 +10563,14 @@ static int convert(const char* from, const char* to, static const char *utf16be = "UTF-16BE"; static iconv_t u8tou16_cd = (iconv_t)-1; static iconv_t u16tou8_cd = (iconv_t)-1; +static pthread_mutex_t o_unac_mutex; +static int unac_mutex_is_init; +// Call this or take your chances with the auto init. +void unac_init_mt() +{ + pthread_mutex_init(&o_unac_mutex, 0); + unac_mutex_is_init = 1; +} /* * Convert buffer containing string encoded in charset into @@ -10576,6 +10583,7 @@ static int convert(const char* from, const char* to, const char* in, size_t in_length, char** outp, size_t* out_lengthp) { + int ret = -1; iconv_t cd; char* out; size_t out_remain; @@ -10584,6 +10592,15 @@ static int convert(const char* from, const char* to, int from_utf16, from_utf8, to_utf16, to_utf8, u8tou16, u16tou8; const char space[] = { 0x00, 0x20 }; + /* Note: better call explicit unac_init_mt() before starting threads than + rely on this. + */ + if (unac_mutex_is_init == 0) { + pthread_mutex_init(&o_unac_mutex, 0); + unac_mutex_is_init = 1; + } + pthread_mutex_lock(&o_unac_mutex); + if (!strcmp(utf16be, from)) { from_utf8 = 0; from_utf16 = 1; @@ -10614,7 +10631,7 @@ static int convert(const char* from, const char* to, /* *outp still valid, no freeing */ if(debug_level >= UNAC_DEBUG_LOW) DEBUG("realloc %d bytes failed\n", out_size+1); - return -1; + goto out; } } else { /* +1 for null */ @@ -10622,7 +10639,7 @@ static int convert(const char* from, const char* to, if(out == 0) { if(debug_level >= UNAC_DEBUG_LOW) DEBUG("malloc %d bytes failed\n", out_size+1); - return -1; + goto out; } } out_remain = out_size; @@ -10631,7 +10648,7 @@ static int convert(const char* from, const char* to, if (u8tou16) { if (u8tou16_cd == (iconv_t)-1) { if((u8tou16_cd = iconv_open(to, from)) == (iconv_t)-1) { - return -1; + goto out; } } else { iconv(u8tou16_cd, 0, 0, 0, 0); @@ -10640,7 +10657,7 @@ static int convert(const char* from, const char* to, } else if (u16tou8) { if (u16tou8_cd == (iconv_t)-1) { if((u16tou8_cd = iconv_open(to, from)) == (iconv_t)-1) { - return -1; + goto out; } } else { iconv(u16tou8_cd, 0, 0, 0, 0); @@ -10648,7 +10665,7 @@ static int convert(const char* from, const char* to, cd = u16tou8_cd; } else { if((cd = iconv_open(to, from)) == (iconv_t)-1) { - return -1; + goto out; } } @@ -10682,7 +10699,7 @@ static int convert(const char* from, const char* to, if(errno == E2BIG) /* fall thru to the E2BIG case below */; else - return -1; + goto out; } else { /* The offending character was replaced by a SPACE, skip it. */ in += 2; @@ -10691,7 +10708,7 @@ static int convert(const char* from, const char* to, break; } } else { - return -1; + goto out; } case E2BIG: { @@ -10711,7 +10728,7 @@ static int convert(const char* from, const char* to, DEBUG("realloc %d bytes failed\n", out_size+1); free(saved); *outp = 0; - return -1; + goto out; } } out = out_base + length; @@ -10719,7 +10736,7 @@ static int convert(const char* from, const char* to, } break; default: - return -1; + goto out; break; } } @@ -10732,7 +10749,10 @@ static int convert(const char* from, const char* to, *out_lengthp = out - out_base; (*outp)[*out_lengthp] = '\0'; - return 0; + ret = 0; +out: + pthread_mutex_unlock(&o_unac_mutex); + return ret; } int unacmaybefold_string(const char* charset, diff --git a/unac/unac.h b/unac/unac.h index c447c8a1..83fb1fea 100644 --- a/unac/unac.h +++ b/unac/unac.h @@ -113,6 +113,9 @@ int unacfold_string(const char* charset, const char* in, size_t in_length, char** out, size_t* out_length); +/* To be called before starting threads in mt programs */ +void unac_init_mt(); + /* * Return unac version number. */ diff --git a/website/BUGS.html b/website/BUGS.html index a07428ce..d067689e 100644 --- a/website/BUGS.html +++ b/website/BUGS.html @@ -38,7 +38,19 @@

recoll 1.16

    - + +
  • The recoll GUI program sometimes crashes when + running a query while the indexing thread is active. + Possible workarounds:
    +
      +
    • (Recommended) Use the command + line recollindex program to perform indexing + (usually just type "recollindex" in a console, or see "man + recollindex").
      • +
    • Do not run queries in recoll while the indexing thread + is running (as indicated in the bottom status line).
      • +
    +
  • Cancelling a preview in the GUI will also cancel the indexing thread if it is running.
    • diff --git a/website/download.html b/website/download.html index ea819dc8..83490d82 100644 --- a/website/download.html +++ b/website/download.html @@ -53,6 +53,27 @@

    The current version is 1.16.0. Release notes.

    +
    +

    Notice for 1.16.0: the + recoll GUI program sometimes crashes when running a query + while the indexing thread is active. I can reproduce the problem + and I am working on a correction. Meanwhile, there are two possible + workarounds:
    +

      +
    • (Recommended) Use the command + line recollindex program to perform indexing + (usually just type "recollindex" in a console, or see "man + recollindex").
      • +
    • Do not run queries in recoll while the indexing thread + is running (as indicated in the bottom status line).
      • +
    + If the workaround fails or you experience other kinds of + crashes with either recoll or recollindex, and + want to help, please follow the instructions on + + this wiki page.

    +
    +

    The download page for Recoll 1.15 is still available.

    @@ -204,21 +225,26 @@ Xapian, Recoll and kio-recoll. These were built from the latest versions, - for a set of Ubuntu series. You just need to add the - PPAs to your system software sources (the instructions are on + for a set of Ubuntu series.

    + +

    Ubuntu 10.04 (lucid) and later versions just need the Recoll + PPA. Older versions also needed a backport for Xapian + (xapian-backports/xapian-1.2).

    + +

    Just add the + PPA to your system software sources (the instructions are on the PPA page or here), and you can then use the normal package - manager to install or update Recoll and Xapian. For Ubuntu versions - from 9.10 (Karmic), only two commands are needed: + manager to install or update Recoll. For Ubuntu versions + after 9.10 (Karmic), only one command is needed: 

    
-sudo add-apt-repository ppa:xapian-backports/xapian-1.2
 sudo add-apt-repository ppa:recoll-backports/recoll-1.15-on

    -

    For Ubuntu 9.04 (Jaunty) and older, to avoid - messages about signature errors, you may have to explicitely import the +

    For Ubuntu 9.04 (Jaunty) and older, + you may have to explicitely import the Recoll and Xapian public keys: 

    
 gpg --keyserver keyserver.ubuntu.com --recv 9DA85604
 gpg --export --armor 9DA85604 | sudo apt-key add -
@@ -390,7 +416,11 @@ I now use the OpenSUSE build service to create Recoll OpenSUSE packages.
       
    Updated 1.16 translations that became available after the
 	release:
    
 
-      
    None for now :(
    
+      
    Czech, thanks to Pavel !
+	recoll_cs.ts
+	recoll_cs.qm
+	
     
+