Swift: add possibility to run the extractor under an env-specified tool

if `CODEQL_EXTRACTOR_SWIFT_RUN_UNDER` env variable is set, and either
* `CODEQL_EXTRACTOR_SWIFT_RUN_UNDER_FILTER` is not set, or
* it is set to a regexp matching any substring of the extractor call
then the extractor process is substituted with the command (and possibly
options) stated in `CODEQL_EXTRACTOR_SWIFT_RUN_UNDER`, followed by the
system arguments of the extractor itself (which should include the
extractor program itself at the start).

Before calling `exec`, `CODEQL_EXTRACTOR_SWIFT_RUN_UNDER` is unset to
avoid unpleasant loops.

An example usage is to run the extractor under `gdbserver :1234` when
the arguments match a given source file.

Author: redsun82

swift/README.md

@@ -24,7 +24,9 @@ Notice you can run `bazel run :create-extractor-pack` if you already are in the
 
 Using `codeql ... --search-path=swift/extractor-pack` will then pick up the Swift extractor. You can also use
 `--search-path=.`, as the extractor pack is mentioned in the root `codeql-workspace.yml`. Alternatively, you can
-set up the search path in [the per-user CodeQL configuration file](https://codeql.github.com/docs/codeql-cli/specifying-command-options-in-a-codeql-configuration-file/#using-a-codeql-configuration-file).
+set up the search path
+in [the per-user CodeQL configuration file](https://codeql.github.com/docs/codeql-cli/specifying-command-options-in-a-codeql-configuration-file/#using-a-codeql-configuration-file)
+.
 
 ## Code generation
 
@@ -53,3 +55,26 @@ options. This is known to have issues on non-Linux platforms.
 The `CMakeLists.txt` file allows to load the Swift extractor as a CMake project, which allows integration into a wider
 variety of IDEs. Building with CMake also creates a `compile_commands.json` compilation database that can be picked up
 by even more IDEs. In particular, opening the `swift` directory in VSCode should work.
+
+### Debugging codeql database creation
+
+If you want to debug a specific run of the extractor within an integration test or a complex `codeql database create`
+invocation, you can do so using [`gdbserver`][gdbserver] or [`lldb-server`][lldb-server].
+
+[gdbserver]: https://sourceware.org/gdb/onlinedocs/gdb/gdbserver-man.html
+
+[lldb-server]: https://lldb.llvm.org/man/lldb-server.html
+
+For example with `gdbserver`, you can
+
+```bash
+export CODEQL_EXTRACTOR_SWIFT_RUN_UNDER="gdbserver :1234"
+export CODEQL_EXTRACTOR_SWIFT_RUN_UNDER_FILTER="SomeSwiftSource\.swift"  # can be any regex matching extractor args
+```
+
+before starting the database extraction, and when that source is encountered the extractor will be run under
+a `gdbserver` instance listening on port 1234. You can then attach to the running debugging server from `gdb` or your
+IDE. Please refer to your IDE's instructions for how to set up remote debugging.
+
+*Note for CLion users*: there is an issue for which breakpoints will not work in this mode if you are opening the
+workspace using the Intellij Bazel plugin. For the time being the workaround is to use the CMake project instead.

swift/extractor/main.cpp

@@ -5,6 +5,8 @@
 #include <vector>
 #include <string>
 #include <iostream>
+#include <regex>
+#include <unistd.h>
 
 #include <swift/Basic/LLVMInitialize.h>
 #include <swift/FrontendTool/FrontendTool.h>
@@ -51,7 +53,51 @@ static void lockOutputSwiftModuleTraps(const codeql::SwiftExtractorConfiguration
   }
 }
 
+// if `CODEQL_EXTRACTOR_SWIFT_RUN_UNDER` env variable is set, and either
+// * `CODEQL_EXTRACTOR_SWIFT_RUN_UNDER_FILTER` is not set, or
+// * it is set to a regexp matching any substring of the extractor call
+// then the running process is substituted with the command (and possibly
+// options) stated in `CODEQL_EXTRACTOR_SWIFT_RUN_UNDER`, followed by `argv`.
+// Before calling `exec`, `CODEQL_EXTRACTOR_SWIFT_RUN_UNDER` is unset to avoid
+// unpleasant loops.
+// An example usage is to run the extractor under `gdbserver :1234` when the
+// arguments match a given source file.
+void checkToRunUnderTool(int argc, char* const* argv) {
+  auto runUnder = getenv("CODEQL_EXTRACTOR_SWIFT_RUN_UNDER");
+  if (runUnder == nullptr) {
+    return;
+  }
+  auto runUnderFilter = getenv("CODEQL_EXTRACTOR_SWIFT_RUN_UNDER_FILTER");
+  if (runUnderFilter != nullptr) {
+    assert(argc > 0);
+    std::string call = argv[0];
+    for (auto i = 0; i < argc; ++i) {
+      call += ' ';
+      call += argv[i];
+    }
+    std::regex filter{runUnderFilter, std::regex_constants::basic | std::regex_constants::nosubs};
+    if (!std::regex_search(call, filter)) {
+      return;
+    }
+  }
+  std::vector<char*> args;
+  for (auto word = std::strtok(runUnder, " "); word != nullptr; word = std::strtok(nullptr, " ")) {
+    args.push_back(word);
+  }
+  if (args.empty()) {
+    return;
+  }
+  for (auto i = 0; i < argc; ++i) {
+    args.push_back(argv[i]);
+  }
+  args.push_back(nullptr);
+  unsetenv("CODEQL_EXTRACTOR_SWIFT_RUN_UNDER");
+  execvp(args[0], args.data());
+}
+
 int main(int argc, char** argv) {
+  checkToRunUnderTool(argc, argv);
+
   if (argc == 1) {
     // TODO: print usage
     return 1;