From 34728dc6e5b868a7b6b64bcc4e1a3a94c1af03845a8698589212005c2d41c0cb Mon Sep 17 00:00:00 2001 From: Peter Simons Date: Tue, 15 Mar 2022 08:17:48 +0000 Subject: [PATCH] Accepting request 961811 from home:coolo:branches:utilities This time I branched all staging failures to make sure it's the last one - A more correct approach to creating symlinks (old bug actually): Do not link the files as given by fdupes, but turn them into relative links (it works by chance if given a buildroot, but fails if running on a subdirectory) - Support multiple directories given (as glob to the macro) OBS-URL: https://build.opensuse.org/request/show/961811 OBS-URL: https://build.opensuse.org/package/show/utilities/fdupes?expand=0&rev=25 --- fdupes.changes | 9 ++++ fdupes_wrapper.cpp | 102 ++++++++++++++++++++++++++++++++++----------- macros.fdupes | 2 +- 3 files changed, 87 insertions(+), 26 deletions(-) diff --git a/fdupes.changes b/fdupes.changes index 89cd3df..4012f13 100644 --- a/fdupes.changes +++ b/fdupes.changes @@ -1,3 +1,12 @@ +------------------------------------------------------------------- +Tue Mar 15 07:41:35 UTC 2022 - Stephan Kulow + +- A more correct approach to creating symlinks (old bug actually): + Do not link the files as given by fdupes, but turn them into + relative links (it works by chance if given a buildroot, but + fails if running on a subdirectory) +- Support multiple directories given (as glob to the macro) + ------------------------------------------------------------------- Mon Mar 14 13:44:54 UTC 2022 - Stephan Kulow diff --git a/fdupes_wrapper.cpp b/fdupes_wrapper.cpp index 9972152..1ef481a 100644 --- a/fdupes_wrapper.cpp +++ b/fdupes_wrapper.cpp @@ -20,10 +20,65 @@ #include #include #include +#include + +using namespace std; typedef std::map> dups_map; typedef std::pair nlink_pair; +vector split_paths(const string& path) +{ + string token; + vector paths; + stringstream ss(path); + while (getline(ss, token, '/')) { + if (token == "..") { + paths.pop_back(); + } else if (token != "." || ss.eof()) { + paths.push_back(token); + } + } + return paths; +} + +string merge_paths(vector paths) +{ + string path; + for (const auto& s : paths) { + if (s.empty()) + continue; + if (!path.empty()) + path += "/"; + path += s; + } + + return path; +} + +string relative(const string& p1, const string& p2) +{ + vector paths1 = split_paths(p1); + paths1.pop_back(); + vector paths2 = split_paths(p2); + vector paths; + vector::const_iterator it1 = paths1.begin(); + vector::const_iterator it2 = paths2.begin(); + // first remove the common parts + while (it1 != paths1.end() && *it1 == *it2) { + it1++; + it2++; + } + for (; it1 != paths1.end(); ++it1) { + paths.push_back(".."); + } + for (; it2 != paths2.end(); ++it2) { + paths.push_back(*it2); + } + + return merge_paths(paths); +} + bool cmp_nlink(const nlink_pair& a, const nlink_pair& b) { return a.second > b.second; @@ -61,6 +116,14 @@ void link_file(const std::string& file, const std::string& target, bool symlink) } } +std::string target_for_link(string target, const std::string &file, bool symlink) +{ + if (!symlink) // hardlinks don't care + return target; + + return relative(file, target); +} + void handle_dups(const dups_map& dups, const std::string& buildroot, bool symlink) { // all are hardlinks to the same data @@ -70,14 +133,11 @@ void handle_dups(const dups_map& dups, const std::string& buildroot, bool symlin sort_by_count(dups, sorted); auto inodes = sorted.begin(); std::string target = dups.at(*inodes).front(); - if (symlink) { - target.replace(0, buildroot.length(), ""); - } for (++inodes; inodes != sorted.end(); ++inodes) { const std::vector files = dups.at(*inodes); for (auto it = files.begin(); it != files.end(); ++it) { - link_file(*it, target, symlink); + link_file(*it, target_for_link(target, *it, symlink), symlink); } } } @@ -85,7 +145,7 @@ void handle_dups(const dups_map& dups, const std::string& buildroot, bool symlin int main(int argc, char** argv) { bool symlink = false; - std::string root; + std::vector roots; std::string buildroot; while (1) { int result = getopt(argc, argv, "sb:"); @@ -95,32 +155,22 @@ int main(int argc, char** argv) case 's': symlink = true; break; - case 'b': - buildroot = optarg; - break; default: /* unknown */ break; } } - if (buildroot.empty()) { - if (symlink) { - std::cerr << "Missing -b argument to remove bootroot from symlink targets"; - return 1; - } - // eliminate final slash from directory argument - if (buildroot.back() == '/') { - buildroot.pop_back(); + while (optind < argc) { + std::string root = argv[optind++]; + if (root.front() != '/') { + char buffer[PATH_MAX]; + root = std::string(getcwd(buffer, PATH_MAX)) + '/' + root; } + roots.push_back(root); } - if (optind < argc) { - root = argv[optind++]; - } else { + + if (roots.empty()) { std::cerr << "Missing directory argument."; } - if (optind < argc) { - std::cerr << "Too many arguments."; - return 1; - } /* fdupes options used: -q: hide progress indicator -p: don't consider files with different owner/group or permission bits as duplicates @@ -129,13 +179,15 @@ int main(int argc, char** argv) -r: follow subdirectories -H: also report hard links as duplicates */ - std::string command = "fdupes -q -p -n -o name"; + std::string command = "fdupes -q -p -r -n -o name"; if (!symlink) { /* if we create symlinks, avoid looking at hard links being duplicated. This way fdupes is faster and won't break them up anyway */ command += " -H"; } - command += " -r '" + root + "'"; + for (auto it = roots.begin(); it != roots.end(); ++it) { + command += " '" + *it + "'"; + } FILE* pipe = popen(command.c_str(), "r"); if (!pipe) { throw std::runtime_error("popen() failed!"); diff --git a/macros.fdupes b/macros.fdupes index fd514d7..be95009 100644 --- a/macros.fdupes +++ b/macros.fdupes @@ -1 +1 @@ -%fdupes /usr/lib/rpm/fdupes_wrapper -b %{buildroot} +%fdupes /usr/lib/rpm/fdupes_wrapper