gdbdump 0.9.2 → 0.9.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 4404278dd2a252a8c175717c167a796adee717c6
-  data.tar.gz: 5b6e430c94796dab5eab7db5933c83757455c295
+  metadata.gz: 47f96b6948a61ee96134de40f22548c97d265549
+  data.tar.gz: 406b7e0a333f356278bc813082cbffc599f7fcc6
 SHA512:
-  metadata.gz: eb73707296f26c46c4c8e8def9a9cab0584e27c6f5c14d62b8b549cb90ca047ccaa8bd17c1dffcefa0063c5e46484f4ab0fc0f3f8d241093e5f38d5eb3136a25
-  data.tar.gz: 6c6ead8c0f79e04ace82dc5dc68fca17fdce94ece83a346110bab569505415b79308b6dd8c3bd281cb97ef4c7bbdcd3184a3f4a6df8d01942650d73f7f50237f
+  metadata.gz: 73adc0c34ac5e358de859b99be87e59da5558cce6fbcc1ba5ffc01d366d1c0d3fe27abe82afc1b2966dccd6b9165e09932147402a09f29172b7345fc30a0e4b3
+  data.tar.gz: eea4efc06c19748d84b626b6593a2afcb89ab9876f5e44c3686658bf222ce9f30d2a875de4efd4ff3db78373b5a5b532655cee35a21db5ea013d733794fe5486

data/CHANGELOG.md

@@ -1,3 +1,9 @@
+# 0.9.3
+
+Enhancements:
+
+* Trivial enhancements of error messages
+
 # 0.9.2
 
 Enhancements:

data/README.md

@@ -21,7 +21,7 @@ It was verfied that gdbdump works with ruby executables built by [rbenv/ruby-bui
 ```
 Usage: gdbdump [options] pid
     -d, --[no-]debug                 print debug log (default: false)
-    -x, --gdbinit FILE               path to ruby trunk's .gdbinit (default: some of ruby trunk's .gdbinit is pre-bundle in this gem)
+    -x, --gdbinit FILE               path to ruby repo's .gdbinit (default: some of ruby repo's .gdbinit is pre-bundle in this gem)
         --gdb PATH                   path to gdb command (default: gdb)
         --ruby PATH                  path to ruby which the attached process uses (default: get from /proc/[PID]/exe)
 ```
@@ -30,7 +30,7 @@ Usage: gdbdump [options] pid
 
 Default supported ruby versions: 2.1.x - 2.4.x
 
-Ruby trunk's [.gdbinit](https://github.com/ruby/ruby/blob/trunk/.gdbinit) file defines useful helper functions and it is maintained by ruby core team. `gdbdump` uses it.
+Ruby repo's [.gdbinit](https://github.com/ruby/ruby/blob/trunk/.gdbinit) file defines useful helper functions and it is maintained by ruby core team. `gdbdump` uses it.
 Some versions written as above of .gdbinit are bundled in this gem, but if you want to use `gdbdump` for older or newer ruby versions:
 
 1. Download .gdbinit from ruby repo like [ruby/2.4.1/.gdbinit](https://github.com/ruby/ruby/blob/v2_4_1/.gdbinit), and specify with `-x` option
@@ -76,11 +76,11 @@ loop.rb:11:in `block in <main>'
   * You can print C level backtrace with raw gdb, of course
 * [sigdump](https://github.com/frsyuki/sigdump)
   * sigdump enables to print ruby level backtrace with sending CONT signal to living ruby process.
-  * The ruby process must pre-install `sigdump` gem and `require 'sigdump/setup'` unlike gdbdump.
+  * The ruby process **must pre-install** `sigdump` gem and `require 'sigdump/setup'` unlike gdbdump.
   * sigdump prints backtrace in signal handler, so blocks main thread, but other threads still work unlike gdbdump.
 * [gdbruby](https://github.com/gunyarakun/gdbruby)
   * gdbruby enables to print C level and ruby level backtrace of living ruby process and core file.
-  * gdbruby must follow changes of C level interfaces of CRuby to get backtrace of the core file, it rises fragility that it will be broken on newer ruby versions.
+  * gdbruby **must follow changes of C level interfaces of CRuby** to get backtrace of the core file, it rises fragility that it will be broken on newer ruby versions.
   * gdbruby stops the process during printing backtrace as gdbdump, but it supports also core file. Using `gcore` command to get core file, it would be possible to analyze without stopping the process.
 
 ## Development
@@ -89,6 +89,70 @@ After checking out the repo, run `bin/setup` to install dependencies. Then, run
 
 To install this gem onto your local machine, run `bundle exec rake install`. To release a new version, update the version number in `version.rb`, and then run `bundle exec rake release`, which will create a git tag for the version, push git commits and tags, and push the `.gem` file to [rubygems.org](https://rubygems.org).
 
+### Investigation
+
+Belows are my investigation notes about how to print C level and Ruby level backtrace via gdb.
+
+#### (1) rb_print_backtrace() and rb_backtrace()
+
+CRuby itself has [rb_print_backtrace](https://github.com/ruby/ruby/blob/26864584d269b6141a27c783cf8b751c067c7dbe/vm_dump.c#L671) function to print C level backtrace, and [rb_backtrace](https://github.com/ruby/ruby/blob/eb59047e2aabb050b23061d513f7b89dc2905670/vm_backtrace.c#L770) function to print Ruby level backtrace.
+I first tried to use them.
+
+However, they print a backtrace of only current thread. Furthermore, rb_print_backtrace supports printing outputs to only STDERR.
+
+```ruby
+def print_backtrace
+  run do |gdb|
+    gdb.cmd_exec('call write(2, "== c backtrace ==\n", 18)')
+    gdb.cmd_exec('call rb_print_backtrace()')
+    gdb.cmd_exec('call write(2, "== ruby backtrace ==\n", 21)')
+    gdb.cmd_exec('call rb_backtrace()')
+  end
+end
+```
+
+#### (2) bt and rb_eval_string()
+
+I secondly tried to use GDB's `info threads` and `bt` command to print C level backtrace. This was fine.
+
+I also tried to print Ruby level backtrace by calling [rb_eval_string](https://github.com/ruby/ruby/blob/44396dbe123511678710cfb21223c954b9ceaafb/vm_eval.c#L1474) with Ruby codes as:
+
+```ruby
+def ruby_backtrace_code
+  code = +%Q[File.open('#{dumpfile}', 'a') {|f|]
+  code << %q[
+    Thread.list.each {|th|
+      f.write %Q[  Thread #{th} status=#{th.status} priority=#{th.priority}\n]
+      th.backtrace.each {|bt|
+        f.write %Q[      #{bt}\n]
+      }
+    }
+  }]
+  code.split("\n").map(&:strip).join('; ')
+end
+
+def print_backtrace
+  run do |gdb|
+    gdb.cmd_exec(%Q[call rb_eval_string("#{ruby_backtrace_code}")])
+  end
+end
+```
+
+However, the debugee (target) process got stuck after `call rb_eval_string`. It seemed the ruby process will be broken if ruby codes are executed via gdb.
+
+#### (3) rb_ps from ruby trunk's .gdbinit
+
+I thirdly tried to use `rb_ps` function defined in [.gdbinit](https://github.com/ruby/ruby/blob/44396dbe123511678710cfb21223c954b9ceaafb/.gdbinit#L983) of ruby repository to print C level and Ruby level backtrace of all threads.
+
+Since `.gdbinit` is maintained by ruby core team, I do not need to follow changes of C level interfaces of CRuby as long as I follow `.gdbinit`.
+
+The drawback of this way is that
+
+1. I have to follow changes of `.gdbinit` in newer ruby versions
+2. `rb_ps` takes somewhat long time like 2.6 seconds to print backtrace.
+
+Currently, I am taking this way.
+
 ## Contributing
 
 Bug reports and pull requests are welcome on GitHub at https://github.com/sonots/gdbdump. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the [Contributor Covenant](http://contributor-covenant.org) code of conduct.

data/lib/gdbdump/cli.rb

@@ -24,7 +24,7 @@ class Gdbdump
       op.on('-d', '--[no-]debug', "print debug log (default: #{opts[:debug]})") {|v|
         opts[:debug] = v
       }
-      op.on('-x', '--gdbinit FILE', "path to ruby trunk's .gdbinit (default: some of ruby trunk's .gdbinit is pre-bundle in this gem)") {|v|
+      op.on('-x', '--gdbinit FILE', "path to ruby repo's .gdbinit (default: some of ruby repo's .gdbinit are pre-bundle in this gem)") {|v|
         opts[:gdbinit] = v
       }
       op.on('--gdb PATH', "path to gdb command (default: gdb)") {|v|

data/lib/gdbdump/gdb.rb

@@ -13,8 +13,12 @@ class Gdbdump
       @pid = pid.to_s
       @debug = debug
       @gdb = gdb || 'gdb'
-      @ruby = ruby || Procfs.new(@pid).exe
-      @gdbinit = gdbinit || File.join(ROOT, 'vendor', 'ruby', ruby_minor_version, 'gdbinit')
+      @ruby = (ruby || Procfs.new(@pid).exe).tap do |path|
+        raise "ruby #{path} is not accessible" unless File.executable?(path)
+      end
+      @gdbinit = (gdbinit || File.join(ROOT, 'vendor', 'ruby', ruby_minor_version, 'gdbinit')).tap do |path|
+        raise "gdbinit #{path} is not readable" unless File.readable?(path)
+      end
       @exec_options = [SUDO_CMD, @gdb, '-silent', '-nw', '-x', @gdbinit, @ruby, @pid]
     end
 

data/lib/gdbdump/version.rb

@@ -1,3 +1,3 @@
 class Gdbdump
-  VERSION = "0.9.2"
+  VERSION = "0.9.3"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: gdbdump
 version: !ruby/object:Gem::Version
-  version: 0.9.2
+  version: 0.9.3
 platform: ruby
 authors:
 - Naotoshi Seo
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2017-06-02 00:00:00.000000000 Z
+date: 2017-06-04 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler