Fix a crash in option parsing.

[aprantl]

Summary

The call to getopt_long didn't handle the case where the *last* option had an argument missing.

[JDevlieghere]

LGTM!

[clayborg]

nice

[stella.stamenova]

This test is extremely flaky on Windows - it looks like it passes on the bot less than 10% of the time: http://lab.llvm.org:8011/buildslaves/win-py3-buildbot.

[mgorny]

The added test broke NetBSD buildbot: http://lab.llvm.org:8011/builders/netbsd-amd64/builds/109/steps/run%20unit%20tests/logs/FAIL%3A%20LLDB%3A%3ATestProcessAttach.test

Apparently this doesn't work as expected here:

$ /home/mgorny/llvm-project/build-rel-master/bin/lldb --no-lldbinit -S /home/mgorny/llvm-project/llvm/tools/lldb/lit/lit-lldb-init -x -b -S /home/mgorny/llvm-project/lldb/lit/Driver/Inputs/process_attach_pid.in                                                                          
(lldb) command source -s 0 '/home/mgorny/llvm-project/llvm/tools/lldb/lit/lit-lldb-init'                                                          
Executing commands in '/home/mgorny/llvm-project/llvm/tools/lldb/lit/lit-lldb-init'.                                                              
(lldb) # LLDB init file for the LIT tests.
(lldb) settings set symbols.enable-external-lookup false
(lldb) settings set interpreter.echo-comment-commands false
(lldb) command source -s 0 '/home/mgorny/llvm-project/lldb/lit/Driver/Inputs/process_attach_pid.in'                                               
Executing commands in '/home/mgorny/llvm-project/lldb/lit/Driver/Inputs/process_attach_pid.in'.                                                   
(lldb) process attach --pid
error: unknown or ambiguous option

[mgorny]

@stella.stamenova could you try running the lldb invocation manually like I did? I'm wondering if you're getting the same issue.

[stella.stamenova]

In D63110#1551572, @mgorny wrote:

@stella.stamenova could you try running the lldb invocation manually like I did? I'm wondering if you're getting the same issue.

It is the same. The interesting thing is that it passes maybe 1/10 of the time on the Buildbot.

[labath]

BTW, the new version was now failing on linux for a change.

I believe r364317 ought to implement this in a way that is compatible with all getopt implementations that we are using...

[aprantl]

Thanks!

lldb/trunk/lit/Driver/Inputs/process_attach_pid.in

+process attach --pid

lldb/trunk/lit/Driver/TestProcessAttach.test

+# RUN: %lldb -x -b -S %S/Inputs/process_attach_pid.in 2>&1 | FileCheck %s
+# CHECK: requires an argument

lldb/trunk/source/Interpreter/Options.cpp

llvm::Expected<Args> Options::Parse(const Args &args,
   std::vector<char *> argv = GetArgvForParsing(args);
   std::unique_lock<std::mutex> lock;
   OptionParser::Prepare(lock);
   int val;
   while (true) {
     int long_options_index = -1;
     val = OptionParser::Parse(argv.size(), &*argv.begin(), sstr.GetString(),
                               long_options, &long_options_index);
+
+    if ((size_t)OptionParser::GetOptionIndex() > argv.size()) {
+      error.SetErrorStringWithFormat("option requires an argument");
+      break;
+    }
+
     if (val == -1)
       break;
 
     // Did we get an error?
     if (val == '?') {
       error.SetErrorStringWithFormat("unknown or ambiguous option");
       break;
     }