Improve code structure and fuctionality

README.md

@@ -1,18 +1,45 @@
 # bundler
 
-A tool to bundle your binaries, useful for building small Docker containers or AppImages.
+A tool that:
+- :deciduous_tree: turns `ldd` into a fancy tree
+- :point_up: explains why `ldd` finds shared libraries and why not
+- :package: optionally deploys relevant executables and dependencies into a single directory
 
-Current functionality:
-- [x] Add executables with `-e` and libraries via `-l`
-- [x] Walks the dependency tree like `ld.so` (handles `RPATH`, `RUNPATH` and `LD_LIBRARY_PATH` correctly).
-- [x] Uses `/etc/ld.so.conf` or any custom conf file via `-ldconf /path/to/ld.so.conf`
-- [x] Skips blacklisted dependencies (and their dependencies) such as libc.so and libstdc++.so.
-- [x] Deploy binaries and rewrite their `RUNPATH`s\*
-- [ ] i386 (currently it's hardcoded to only deploy x86_64)
-- [ ] Ship `chrpath`
+## Example 1: listing the dependencies of an executable
+
+![example](doc/screenshot.png)
+
+## Example 2: deploying binaries + dependencies into a folder:
+```bash
+$ bundler -e $(which bundler) -d bundler.bundle
+Dependency tree
+bundler
+├── libcppglob.so.1 [direct]
+│   ├── libstdc++.so.6 (skipped) [ld.so.conf]
+│   ├── libgcc_s.so.1 (skipped) [ld.so.conf]
+│   └── libc.so.6 (skipped) [ld.so.conf]
+├── libstdc++.so.6 (skipped) [ld.so.conf]
+├── libgcc_s.so.1 (skipped) [ld.so.conf]
+└── libc.so.6 (skipped) [ld.so.conf]
+
+Deploying to "bundler.bundle/usr"
+"/home/.../Documents/projects/bundler/build/bundler" => "bundler.bundle/usr/bin/bundler"
+"/home/.../Documents/projects/bundler/build/lib/libcppglob.so.1.1.0" => "bundler.bundle/usr/lib/libcppglob.so.1.1.0"
+  creating symlink "bundler.bundle/usr/lib/libcppglob.so.1"
+
+$ tree bundler.bundle/
+bundler.bundle/
+└── usr
+    ├── bin
+    │   └── bundler
+    └── lib
+        ├── libcppglob.so.1 -> libcppglob.so.1.1.0
+        └── libcppglob.so.1.1.0
 
-\* Note: `patchelf` seems to be very broken software, so instead I'm using `chrpath`, but this will only patch rpaths _when they already exist in the binary_. Therefore you might still need to add the `yourapp/lib` folder to ld's search paths by running `ldconfig yourapp/lib` or setting `LD_LIBRARY_PATH=yourapp/lib`.
+3 directories, 3 files
+```
 
+## Build from source 
 ```bash
 git clone --recursive https://github.com/haampie/bundler.git
 cd bundler
@@ -22,29 +49,13 @@ cmake -DCMAKE_BUILD_TYPE=Release ..
 make
 ```
 
-```bash
-$ ./bundler -e bundler -d Bundler.app
-Dependency tree
-bundler
-|--libcppglob.so.1
-|  |--libstdc++.so.6 (excluded)
-|  |--libgcc_s.so.1 (excluded)
-|  |--libc.so.6 (excluded)
-|--libstdc++.so.6 (excluded)
-|--libgcc_s.so.1 (excluded)
-|--libc.so.6 (excluded)
-
-Deploying to "Bundler.app/usr"
-"/home/harmen/Documents/projects/binary_bundler/build/bundler" => "Bundler.app/usr/bin/bundler"
-"/home/harmen/Documents/projects/binary_bundler/build/lib/libcppglob.so.1.1.0" => "Bundler.app/usr/lib/libcppglob.so.1"
-
-$ tree Bundler
-Bundler.app/
-└── usr
-    ├── bin
-    │   └── bundler
-    └── lib
-        └── libcppglob.so.1
+## Current functionality
+- [x] Add executables with `-e` and libraries via `-l`
+- [x] Walks the dependency tree like `ld.so` (handles `RPATH`, `RUNPATH` and `LD_LIBRARY_PATH` correctly).
+- [x] Uses `/etc/ld.so.conf` or any custom conf file via `-ldconf /path/to/ld.so.conf`
+- [x] Skips blacklisted dependencies (and their dependencies) such as libc.so and libstdc++.so.
+- [x] Deploy binaries and rewrite their `RUNPATH`s\*
+- [x] Ship `chrpath` and `strip`
+- [ ] i386 (currently it's hardcoded to only deploy x86_64)
 
-3 directories, 2 files
-```
+\* Note: `patchelf` seems to be very broken software, so instead I'm using `chrpath`. The downside is `chrpath` will only patch rpaths _when they already exist in the binary_. Therefore you might still need to add the `yourapp/lib` folder to ld's search paths by running `echo /path/to/yourapp.bundle/lib > /etc/ld.so.conf/my_app.conf && ldconfig` or by setting `LD_LIBRARY_PATH=/path/to/yourapp.bundle/lib`.

include/bundler/deploy.hpp

@@ -11,4 +11,4 @@ namespace fs = std::filesystem;
 std::string repeat(std::string_view input, size_t num);
 
 // Copy binaries over, change their rpath if they have it, and strip them
-void deploy(std::vector<Elf> const &deps, fs::path const &bin, fs::path const &lib, fs::path const &strip, fs::path const &chrpath);
+void deploy(std::vector<Elf> const &deps, fs::path const &bin, fs::path const &lib, fs::path const &chrpath_path, fs::path const &strip_path, bool chrpath, bool strip);

include/bundler/deps.hpp

@@ -10,21 +10,27 @@
 class deps {
 
 public:
-    deps(std::vector<Elf> &&input, std::vector<fs::path> &&search_paths, std::vector<fs::path> && ld_library_paths);
+    deps(std::vector<Elf> &&input, std::vector<fs::path> &&ld_so_conf, std::vector<fs::path> && ld_library_paths, bool verbose);
 
     std::vector<Elf> const &get_deps() const;
 
 private:
-    void explore(Elf const &elf, std::vector<fs::path> &rpaths, size_t depth = 0);
+    void explore(Elf const &elf, std::vector<fs::path> &rpaths, std::vector<bool> &done);
+    void explore(Elf const &elf, std::vector<fs::path> &rpaths);
 
-    std::optional<Elf> try_paths(std::vector<fs::path> const &paths, fs::path const &so);
-    std::optional<Elf> search(fs::path const &search, fs::path const &so);
+    std::optional<Elf> locate(Elf const &parent, fs::path const &so, std::vector<fs::path> const &rpaths, std::vector<fs::path> const &runpaths);
+    std::optional<Elf> locate_directly(Elf const &parent, fs::path const &so);
+    std::optional<Elf> locate_by_search(fs::path const &so, std::vector<fs::path> const &rpaths, std::vector<fs::path> const &runpaths);
+    std::optional<Elf> find_by_paths(fs::path const &so, std::vector<fs::path> const &paths, found_t tag);
 
     size_t m_depth = 0;
     std::vector<Elf> m_top_level;
-    std::vector<fs::path> m_search_paths;
     std::vector<fs::path> m_ld_library_paths;
+    std::vector<fs::path> m_ld_so_conf;
+    std::vector<fs::path> m_default_paths{"/lib", "/usr/lib"};
     std::unordered_set<fs::path, PathHash> m_visited;
 
     std::vector<Elf> m_all_binaries;
+
+    bool m_verbose;
 };

include/bundler/elf.hpp

@@ -10,8 +10,11 @@ namespace fs = std::filesystem;
 
 enum class deploy_t { EXECUTABLE, LIBRARY };
 
+enum class found_t { NONE, DIRECT, RPATH, LD_LIBRARY_PATH, RUNPATH, LD_SO_CONF, DEFAULT_PATHS };
+
 struct Elf {
     deploy_t type;
+    found_t found_via;
     std::string name;
     fs::path abs_path;
     std::vector<fs::path> runpaths;
@@ -34,4 +37,4 @@ fs::path apply_substitutions(fs::path const &rpath, fs::path const &cwd);
 std::vector<fs::path> split_paths(std::string_view raw_path);
 
 // Try to create an elf from a path
-std::optional<Elf> from_path(deploy_t type, fs::path path_str);
+std::optional<Elf> from_path(deploy_t type, found_t found_via, fs::path path_str);

src/deploy.cpp

@@ -4,7 +4,7 @@
 #include <termcolor/termcolor.hpp>
 
 // Copy binaries over, change their rpath if they have it, and strip them
-void deploy(std::vector<Elf> const &deps, fs::path const &bin, fs::path const &lib, fs::path const &strip, fs::path const &chrpath) {
+void deploy(std::vector<Elf> const &deps, fs::path const &bin, fs::path const &lib, fs::path const &chrpath_path, fs::path const &strip_path, bool chrpath, bool strip) {
     for (auto const &elf : deps) {
 
         // Go through all symlinks.
@@ -28,12 +28,16 @@ void deploy(std::vector<Elf> const &deps, fs::path const &bin, fs::path const &l
         auto rpath = (elf.type == deploy_t::EXECUTABLE ? "\\$ORIGIN/../lib" : "\\$ORIGIN");
 
 		// Silently patch the rpath and strip things; let's not care if it fails.
-        std::stringstream chrpath_cmd;
-        chrpath_cmd << chrpath << " -c -r \"" << rpath << "\" " << deploy_path;
-        exec(chrpath_cmd.str().c_str());
+        if (chrpath) {
+            std::stringstream chrpath_cmd;
+            chrpath_cmd << chrpath_path << " -c -r \"" << rpath << "\" " << deploy_path;
+            exec(chrpath_cmd.str().c_str());
+        }
 
-        std::stringstream strip_cmd;
-        strip_cmd << strip << ' ' << deploy_path;
-        exec(strip_cmd.str().c_str());
+        if (strip) {
+            std::stringstream strip_cmd;
+            strip_cmd << strip_path << ' ' << deploy_path;
+            exec(strip_cmd.str().c_str());
+        }
     }
 }