diff options
author | Andrey Lihatskiy <alihatskiy@productengine.com> | 2024-04-29 07:43:28 +0300 |
---|---|---|
committer | Andrey Lihatskiy <alihatskiy@productengine.com> | 2024-04-29 07:56:09 +0300 |
commit | 1b68f71348ecf3983b76b40d7940da8377f049b7 (patch) | |
tree | 2974eddaef130a067c26033d60a59fc790365b3d /indra/newview/llurl.cpp | |
parent | af4ea94efc1999f3b19fd8d643d0331f0b77e265 (diff) |
#824 Process source files in bulk: replace tabs with spaces, convert CRLF to LF, and trim trailing whitespaces as needed
Diffstat (limited to 'indra/newview/llurl.cpp')
-rw-r--r-- | indra/newview/llurl.cpp | 418 |
1 files changed, 209 insertions, 209 deletions
diff --git a/indra/newview/llurl.cpp b/indra/newview/llurl.cpp index 01a81c5f83..2cb7f1a877 100644 --- a/indra/newview/llurl.cpp +++ b/indra/newview/llurl.cpp @@ -1,25 +1,25 @@ -/** +/** * @file llurl.cpp * @brief Text url class * * $LicenseInfo:firstyear=2001&license=viewerlgpl$ * Second Life Viewer Source Code * Copyright (C) 2010, Linden Research, Inc. - * + * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; * version 2.1 of the License only. - * + * * This library 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 * Lesser General Public License for more details. - * + * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA - * + * * Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA * $/LicenseInfo$ */ @@ -30,109 +30,109 @@ LLURL::LLURL() { - init(""); + init(""); } LLURL::LLURL(const LLURL &url) { - if (this != &url) - { - init(url.getFQURL()); - } - else - { - init(""); - } + if (this != &url) + { + init(url.getFQURL()); + } + else + { + init(""); + } } LLURL::LLURL(const char * url) { - init(url); + init(url); } LLURL::~LLURL() { - cleanup(); + cleanup(); } void LLURL::init(const char * url) { - mURI[0] = '\0'; - mAuthority[0] = '\0'; - mPath[0] = '\0'; - mFilename[0] = '\0'; - mExtension[0] = '\0'; - mTag[0] = '\0'; - - char url_copy[MAX_STRING]; /* Flawfinder: ignore */ - - strncpy (url_copy,url, MAX_STRING -1); /* Flawfinder: ignore */ - url_copy[MAX_STRING -1] = '\0'; - - char *parse; - char *leftover_url = url_copy; - S32 span = 0; - - // copy and lop off tag - if ((parse = strchr(url_copy,'#'))) - { - strncpy(mTag,parse+1, LL_MAX_PATH -1); /* Flawfinder: ignore */ - mTag[LL_MAX_PATH -1] = '\0'; - *parse = '\0'; - } - - // copy out URI if it exists, update leftover_url - if ((parse = strchr(url_copy,':'))) - { - *parse = '\0'; - strncpy(mURI,leftover_url, LL_MAX_PATH -1); /* Flawfinder: ignore */ - mURI[LL_MAX_PATH -1] = '\0'; - leftover_url = parse + 1; - } - - // copy out authority if it exists, update leftover_url - if ((leftover_url[0] == '/') && (leftover_url[1] == '/')) - { - leftover_url += 2; // skip the "//" - - span = strcspn(leftover_url, "/"); - strncat(mAuthority,leftover_url,span); /* Flawfinder: ignore */ - leftover_url += span; - } - - if ((parse = strrchr(leftover_url,'.'))) - { - // copy and lop off extension - strncpy(mExtension,parse+1, LL_MAX_PATH -1); /* Flawfinder: ignore */ - mExtension[LL_MAX_PATH -1] = '\0'; - *parse = '\0'; - } - - if ((parse = strrchr(leftover_url,'/'))) - { - parse++; - } - else - { - parse = leftover_url; - } - - // copy and lop off filename - strncpy(mFilename,parse, LL_MAX_PATH -1);/* Flawfinder: ignore */ - mFilename[LL_MAX_PATH -1] = '\0'; - *parse = '\0'; - - // what's left should be the path - strncpy(mPath,leftover_url, LL_MAX_PATH -1); /* Flawfinder: ignore */ - mPath[LL_MAX_PATH -1] = '\0'; - -// LL_INFOS() << url << " decomposed into: " << LL_ENDL; -// LL_INFOS() << " URI : <" << mURI << ">" << LL_ENDL; -// LL_INFOS() << " Auth: <" << mAuthority << ">" << LL_ENDL; -// LL_INFOS() << " Path: <" << mPath << ">" << LL_ENDL; -// LL_INFOS() << " File: <" << mFilename << ">" << LL_ENDL; -// LL_INFOS() << " Ext : <" << mExtension << ">" << LL_ENDL; -// LL_INFOS() << " Tag : <" << mTag << ">" << LL_ENDL; + mURI[0] = '\0'; + mAuthority[0] = '\0'; + mPath[0] = '\0'; + mFilename[0] = '\0'; + mExtension[0] = '\0'; + mTag[0] = '\0'; + + char url_copy[MAX_STRING]; /* Flawfinder: ignore */ + + strncpy (url_copy,url, MAX_STRING -1); /* Flawfinder: ignore */ + url_copy[MAX_STRING -1] = '\0'; + + char *parse; + char *leftover_url = url_copy; + S32 span = 0; + + // copy and lop off tag + if ((parse = strchr(url_copy,'#'))) + { + strncpy(mTag,parse+1, LL_MAX_PATH -1); /* Flawfinder: ignore */ + mTag[LL_MAX_PATH -1] = '\0'; + *parse = '\0'; + } + + // copy out URI if it exists, update leftover_url + if ((parse = strchr(url_copy,':'))) + { + *parse = '\0'; + strncpy(mURI,leftover_url, LL_MAX_PATH -1); /* Flawfinder: ignore */ + mURI[LL_MAX_PATH -1] = '\0'; + leftover_url = parse + 1; + } + + // copy out authority if it exists, update leftover_url + if ((leftover_url[0] == '/') && (leftover_url[1] == '/')) + { + leftover_url += 2; // skip the "//" + + span = strcspn(leftover_url, "/"); + strncat(mAuthority,leftover_url,span); /* Flawfinder: ignore */ + leftover_url += span; + } + + if ((parse = strrchr(leftover_url,'.'))) + { + // copy and lop off extension + strncpy(mExtension,parse+1, LL_MAX_PATH -1); /* Flawfinder: ignore */ + mExtension[LL_MAX_PATH -1] = '\0'; + *parse = '\0'; + } + + if ((parse = strrchr(leftover_url,'/'))) + { + parse++; + } + else + { + parse = leftover_url; + } + + // copy and lop off filename + strncpy(mFilename,parse, LL_MAX_PATH -1);/* Flawfinder: ignore */ + mFilename[LL_MAX_PATH -1] = '\0'; + *parse = '\0'; + + // what's left should be the path + strncpy(mPath,leftover_url, LL_MAX_PATH -1); /* Flawfinder: ignore */ + mPath[LL_MAX_PATH -1] = '\0'; + +// LL_INFOS() << url << " decomposed into: " << LL_ENDL; +// LL_INFOS() << " URI : <" << mURI << ">" << LL_ENDL; +// LL_INFOS() << " Auth: <" << mAuthority << ">" << LL_ENDL; +// LL_INFOS() << " Path: <" << mPath << ">" << LL_ENDL; +// LL_INFOS() << " File: <" << mFilename << ">" << LL_ENDL; +// LL_INFOS() << " Ext : <" << mExtension << ">" << LL_ENDL; +// LL_INFOS() << " Tag : <" << mTag << ">" << LL_ENDL; } void LLURL::cleanup() @@ -142,149 +142,149 @@ void LLURL::cleanup() // Copy assignment LLURL &LLURL::operator=(const LLURL &rhs) { - if (this != &rhs) - { - this->init(rhs.getFQURL()); - } - - return *this; + if (this != &rhs) + { + this->init(rhs.getFQURL()); + } + + return *this; } // Compare bool LLURL::operator==(const LLURL &rhs) const { - if ((strcmp(mURI, rhs.mURI)) - || (strcmp(mAuthority, rhs.mAuthority)) - || (strcmp(mPath, rhs.mPath)) - || (strcmp(mFilename, rhs.mFilename)) - || (strcmp(mExtension, rhs.mExtension)) - || (strcmp(mTag, rhs.mTag)) - ) - { - return FALSE; - } - return TRUE; + if ((strcmp(mURI, rhs.mURI)) + || (strcmp(mAuthority, rhs.mAuthority)) + || (strcmp(mPath, rhs.mPath)) + || (strcmp(mFilename, rhs.mFilename)) + || (strcmp(mExtension, rhs.mExtension)) + || (strcmp(mTag, rhs.mTag)) + ) + { + return FALSE; + } + return TRUE; } bool LLURL::operator!=(const LLURL& rhs) const { - return !(*this == rhs); + return !(*this == rhs); } const char * LLURL::getFQURL() const { - char fqurl[LL_MAX_PATH]; /* Flawfinder: ignore */ - - fqurl[0] = '\0'; - - if (mURI[0]) - { - strncat(fqurl,mURI, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ - strcat(fqurl,":"); /* Flawfinder: ignore */ - if (mAuthority[0]) - { - strcat(fqurl,"//"); /* Flawfinder: ignore */ - } - } - - if (mAuthority[0]) - { - strncat(fqurl,mAuthority, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ - } - - strncat(fqurl,mPath, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ - - strncat(fqurl,mFilename, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ - - if (mExtension[0]) - { - strcat(fqurl,"."); /* Flawfinder: ignore */ - strncat(fqurl,mExtension, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ - } - - if (mTag[0]) - { - strcat(fqurl,"#"); /* Flawfinder: ignore */ - strncat(fqurl,mTag, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ - } - - strncpy(LLURL::sReturnString,fqurl, LL_MAX_PATH -1); /* Flawfinder: ignore */ - LLURL::sReturnString[LL_MAX_PATH -1] = '\0'; - - return(LLURL::sReturnString); + char fqurl[LL_MAX_PATH]; /* Flawfinder: ignore */ + + fqurl[0] = '\0'; + + if (mURI[0]) + { + strncat(fqurl,mURI, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ + strcat(fqurl,":"); /* Flawfinder: ignore */ + if (mAuthority[0]) + { + strcat(fqurl,"//"); /* Flawfinder: ignore */ + } + } + + if (mAuthority[0]) + { + strncat(fqurl,mAuthority, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ + } + + strncat(fqurl,mPath, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ + + strncat(fqurl,mFilename, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ + + if (mExtension[0]) + { + strcat(fqurl,"."); /* Flawfinder: ignore */ + strncat(fqurl,mExtension, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ + } + + if (mTag[0]) + { + strcat(fqurl,"#"); /* Flawfinder: ignore */ + strncat(fqurl,mTag, LL_MAX_PATH - strlen(fqurl) -1); /* Flawfinder: ignore */ + } + + strncpy(LLURL::sReturnString,fqurl, LL_MAX_PATH -1); /* Flawfinder: ignore */ + LLURL::sReturnString[LL_MAX_PATH -1] = '\0'; + + return(LLURL::sReturnString); } const char* LLURL::updateRelativePath(const LLURL &url) { - char new_path[LL_MAX_PATH]; /* Flawfinder: ignore */ - char tmp_path[LL_MAX_PATH]; /* Flawfinder: ignore */ - - char *parse; - - if (mPath[0] != '/') - { - //start with existing path - strncpy (new_path,url.mPath, LL_MAX_PATH -1); /* Flawfinder: ignore */ - new_path[LL_MAX_PATH -1] = '\0'; - strncpy (tmp_path,mPath, LL_MAX_PATH -1); /* Flawfinder: ignore */ - tmp_path[LL_MAX_PATH -1] = '\0'; - - parse = strtok(tmp_path,"/"); - while (parse) - { - if (!strcmp(parse,".")) - { - // skip this, it's meaningless - } - else if (!strcmp(parse,"..")) - { - if ((parse = strrchr(new_path, '/'))) - { - *parse = '\0'; - if ((parse = strrchr(new_path, '/'))) - { - *(parse+1) = '\0'; - } - else - { - new_path[0] = '\0'; - } - } - else - { - strcat(new_path,"../"); /* Flawfinder: ignore */ - } - - } - else - { - strncat(new_path,parse, LL_MAX_PATH - strlen(new_path) -1 ); /* Flawfinder: ignore */ - strcat(new_path,"/"); /* Flawfinder: ignore */ - } - parse = strtok(NULL,"/"); - } - strncpy(mPath,new_path, LL_MAX_PATH -1); /* Flawfinder: ignore */ - mPath[LL_MAX_PATH -1] = '\0'; - } - return mPath; + char new_path[LL_MAX_PATH]; /* Flawfinder: ignore */ + char tmp_path[LL_MAX_PATH]; /* Flawfinder: ignore */ + + char *parse; + + if (mPath[0] != '/') + { + //start with existing path + strncpy (new_path,url.mPath, LL_MAX_PATH -1); /* Flawfinder: ignore */ + new_path[LL_MAX_PATH -1] = '\0'; + strncpy (tmp_path,mPath, LL_MAX_PATH -1); /* Flawfinder: ignore */ + tmp_path[LL_MAX_PATH -1] = '\0'; + + parse = strtok(tmp_path,"/"); + while (parse) + { + if (!strcmp(parse,".")) + { + // skip this, it's meaningless + } + else if (!strcmp(parse,"..")) + { + if ((parse = strrchr(new_path, '/'))) + { + *parse = '\0'; + if ((parse = strrchr(new_path, '/'))) + { + *(parse+1) = '\0'; + } + else + { + new_path[0] = '\0'; + } + } + else + { + strcat(new_path,"../"); /* Flawfinder: ignore */ + } + + } + else + { + strncat(new_path,parse, LL_MAX_PATH - strlen(new_path) -1 ); /* Flawfinder: ignore */ + strcat(new_path,"/"); /* Flawfinder: ignore */ + } + parse = strtok(NULL,"/"); + } + strncpy(mPath,new_path, LL_MAX_PATH -1); /* Flawfinder: ignore */ + mPath[LL_MAX_PATH -1] = '\0'; + } + return mPath; } const char * LLURL::getFullPath() { - strncpy(LLURL::sReturnString,mPath, LL_MAX_PATH -1); /* Flawfinder: ignore */ - LLURL::sReturnString[LL_MAX_PATH -1] = '\0'; - strncat(LLURL::sReturnString,mFilename, LL_MAX_PATH - strlen(LLURL::sReturnString) -1); /* Flawfinder: ignore */ - strcat(LLURL::sReturnString,"."); /* Flawfinder: ignore */ - strncat(LLURL::sReturnString,mExtension, LL_MAX_PATH - strlen(LLURL::sReturnString) -1); /* Flawfinder: ignore */ - return(sReturnString); + strncpy(LLURL::sReturnString,mPath, LL_MAX_PATH -1); /* Flawfinder: ignore */ + LLURL::sReturnString[LL_MAX_PATH -1] = '\0'; + strncat(LLURL::sReturnString,mFilename, LL_MAX_PATH - strlen(LLURL::sReturnString) -1); /* Flawfinder: ignore */ + strcat(LLURL::sReturnString,"."); /* Flawfinder: ignore */ + strncat(LLURL::sReturnString,mExtension, LL_MAX_PATH - strlen(LLURL::sReturnString) -1); /* Flawfinder: ignore */ + return(sReturnString); } const char * LLURL::getAuthority() { - strncpy(LLURL::sReturnString,mAuthority, LL_MAX_PATH -1); /* Flawfinder: ignore */ - LLURL::sReturnString[LL_MAX_PATH -1] = '\0'; - return(sReturnString); + strncpy(LLURL::sReturnString,mAuthority, LL_MAX_PATH -1); /* Flawfinder: ignore */ + LLURL::sReturnString[LL_MAX_PATH -1] = '\0'; + return(sReturnString); } char LLURL::sReturnString[LL_MAX_PATH] = ""; |