[Lldb-commits] [lldb] r296814 - Added a list of outstanding projects that would benefit lldb.
Jim Ingham via lldb-commits
lldb-commits at lists.llvm.org
Thu Mar 2 13:39:27 PST 2017
Author: jingham
Date: Thu Mar 2 15:39:27 2017
New Revision: 296814
URL: http://llvm.org/viewvc/llvm-project?rev=296814&view=rev
Log:
Added a list of outstanding projects that would benefit lldb.
This was a list that I've had kicking around for a while, and would
add to whenever some hallway conversation would bring up another good
idea. It's not doing any good on my desktop, but it might generate
some inspiration here. Please add to this if you have any other
good ideas.
I apologize for the formatting, but if I wait to get it looking
nice it would probably stay on my desktop.
Added:
lldb/trunk/www/projects.html (with props)
Modified:
lldb/trunk/www/sidebar.incl
Added: lldb/trunk/www/projects.html
URL: http://llvm.org/viewvc/llvm-project/lldb/trunk/www/projects.html?rev=296814&view=auto
==============================================================================
--- lldb/trunk/www/projects.html (added)
+++ lldb/trunk/www/projects.html Thu Mar 2 15:39:27 2017
@@ -0,0 +1,379 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1" />
+<link href="style.css" rel="stylesheet" type="text/css" />
+<title>LLDB Projects</title>
+</head>
+
+<body>
+ <div class="www_title">
+ Project ideas for the <strong>LLDB</strong> Debugger
+ </div>
+
+<div id="container">
+ <div id="content">
+ <!--#include virtual="sidebar.incl"-->
+ <div id="middle">
+ <div class="post">
+ <h1 class ="postheader">Goals</h1>
+ <div class="postcontent">
+
+ <p>The following is a mostly unordered set of the ideas for improvements
+ to the LLDB debugger. Some are fairly deep, some would require less
+ effort.
+ </p>
+
+ <ol>
+ <li>
+ Speed up type realization in lldb.
+ <p>
+ The type of problem I'm addressing here is the situation where you are
+ debugging a large program (lldb built with debug clang/swift will do)
+ and you go to print a simple expression, and lldb goes away for 30 seconds.
+ When you sample it, it is always busily churning through all the CU's in the
+ world looking for something. The problem isn't that looking for something in
+ particular is slow, but rather that we somehow turned an bounded search
+ (maybe a subtype of "std::string" into an unbounded search (all things with the
+ name of that subtype.) Or didn't stop when we got a reasonable answer
+ proximate to the context of the search, but let the search leak out globally.
+ And quite likely there are other issues that I haven't guessed yet.
+ But if you end up churning though 3 or 4 Gig of debug info, that's going to be slow
+ no matter how well written your debug reader is...
+ </p>
+ <p>
+ My guess is the work will be more in the general symbol lookup than in the DWARF
+ parser in particular, but it may be a combination of both.
+ </p>
+ <p>
+ As a user debugging a largish program, this is the most obvious lameness of lldb.
+ </p>
+ </li>
+
+ <li>
+ Symbol name completion in the expression parser.
+ <p>
+ This is the other obvious lameness of lldb. You can do:
+ </p>
+<code><pre><tt>
+(lldb) frame var foo.b<TAB>
+</tt></pre></code>
+ <p>
+ and we will tell you it is "foo.bar". But you can't do that in the expression parser.
+ This will require collaboration with the clang/swift folks to get the right extension
+ points in the compiler. And whatever they are, lldb will need use them to tell the
+ compiler about what names are available. It will be important to avoid the pitfalls
+ of #1 where we wander into the entire DWARF world.
+ </p>
+
+ </li>
+ <li>
+ Make a high speed asynchronous communication channel to replace the gdb-remote protocol.
+ <p>
+ All lldb debugging nowadays is done by talking to a debug agent. We used the
+ gdb-remote protocol because that is universal, and good enough, and you have
+ to support it anyway since so many little devices & JTAG's and VM's etc
+ support it. But it is really old, not terribly high performance, and
+ can't really handle sending or receiving messages while the process is
+ supposedly running. It should have compression built in, remove the hand-built
+ checksums and rely on the robust communication protocols we always have nowadays,
+ allow for out-of-order requests/replies, allow for reconnecting to a temporarily
+ disconnected debug session, regularize all of the packet formatting into JSON or
+ BSON or whatever while including a way to do large binary transfers. It must be
+ possible to come up with something faster, and better tunable for the many
+ communications pathways we end up supporting.
+ </p>
+ </li>
+
+ <li>
+ Fix the event handling/process control machinery to support calling SB & Commands
+ everywhere, and to support non-stop debugging
+ <p>
+ There is a fairly ad-hoc system to handle when it is safe to run SB API's and
+ command line commands. This is actually a bit of a tricky problem, since we
+ allow access to the command line and SB API from some funky places in lldb.
+ The Operating System plugins are the most obvious instance, since they get
+ run right after lldb is told by debugserver that the process has stopped, but
+ before it has finished collating the information from the stop for presentation
+ to the higher levels. But breakpoint callbacks have some of the same problems,
+ and other things like the scripted stepping operations and any fancier extension
+ points we want to add to the debugger are going to be hard to implement robustly
+ till we work on a finer-grained and more explicit control over who gets to control
+ the process state.
+ <p>
+ We also won't have any chance of supporting non-stop debugging - which is a useful
+ mode for programs that have a lot of high-priority or real-time worker threads -
+ until we get this sorted out.
+ </p>
+ </li>
+
+ <li>
+ Finish the language abstraction and remove all the unnecessary C++/clang-based API's
+ <p>
+ An important part of making lldb a more useful "debugger toolkit" as opposed to a
+ C/C++/ObjC/Swift debugger is to have a clean abstraction for language support.
+ We did most, but not all, of the physical separation. We need to finish that.
+ And then by force of necessity the API's really look like the interface to a C++
+ type system with a few swift bits added on. How you would go about adding a new
+ language is unclear and much more trouble than it is worth at present. But if
+ we made this nice, we could add a lot of value to other language projects.
+ </p>
+ </li>
+
+ <li>
+ Add some syntax to generate data formatters from type definitions
+ <p>
+ Uses of the data formatters fall into two types. There are data formatters for
+ types where the structure elements pretty much tell you how to present the data,
+ you just need a little expression language to express how to turn them into what
+ the user expects to see. Then there are the ones (like pretty much all our
+ Foundation/AppKit/UIKit formatters) that use deep magic to figure out how the type
+ is actually laid out. The latter are pretty much always going to have to be done by hand.
+ </p>
+ <p>
+ But for the ones where the information is expressed in the fields, it would be great
+ to have a way to express the instructions to produce summaries and children in some
+ form you could embed next to the types and have the compiler produce a byte code
+ form of the instructions and then make that available to lldb along with the library.
+ This isn't as simple as having clang run over the headers and produce something from the
+ types directly. After all, clang has no way of knowing that the interesting thing about
+ a std::vector is the elements that you get by calling size (for the summary) and [] for
+ the elements. But it shouldn't be hard to come up with a generic markup to express this.
+ </p>
+ </li>
+
+ <li>
+ Allow the expression parser to access dynamic type/data formatter information
+ <p>
+ This seems like a smaller one. The symptom is your object is Foo child of Bar, and in
+ the Locals view you see all the fields of Foo, but because the static type of the object
+ is Bar, you can't see any of the fields of Foo. But if you could get this working,
+ you could hijack the mechanism to make the results of the value object summaries/synthetic
+ children available to expressions. And if you can do that, you could add other properties
+ to an object externally (through Python or some other extension point) and then have
+ these also available in the expression parser. You could use this to express invariants
+ for data structures, or other more advanced uses of types in the debugger.
+ </p>
+ <p>
+ Another version of this is to allow access to synthetic children in the expression
+ parser. Otherwise you end up in situations like:
+ </p>
+<code><pre><tt>
+(lldb) print return_a_foo()
+(SomeVectorLikeType) $1 = {
+ [0] = 0
+ [1] = 1
+ [2] = 2
+ [3] = 3
+ [4] = 4
+}
+</tt></pre></code>
+
+ <p>
+ That's good but:
+ </p>
+<code><pre><tt>
+(lldb) print return_a_foo()[2]
+</tt></pre></code>
+ <p>
+ fails because the expression parser doesn't know anything about the array-like nature of
+ SomeVectorLikeType that it gets from the synthetic children.
+ </p>
+ </li>
+
+ <li>
+ Add an extension point in the breakpoint search machinery.
+ <p>
+ This would allow highly customizable, algorithmic breakpoint types, like "break on every
+ use of some particular instruction, or instruction pattern, etc."
+ </p>
+ </li>
+
+ <li>
+ Documentation and better examples
+
+ <p>
+ We need to put the lldb syntax docs in the tutorial somewhere that is more easily accessible.
+ On suggestion is to add non-command based help to the help system, and then have a "help lldb"
+ or "help syntax" type command with this info. Be nice if the non-command based help could
+ be hierarchical so you could make topics.
+ </p>
+
+ <p>
+ There's a fair bit of docs about the SB API's, but it is spotty. Some classes are
+ well documented in the Python "help (lldb.SBWhatever)" and some are not.
+ </p>
+
+
+ <p>
+ We need more conceptual docs. And we need more examples. And we could provide a
+ clean pluggable example for using LLDB standalone from Python. The process_events.py
+ is a start of this, but it just handles process events, and it is really a quick sketch
+ not a polished expandable proto-tool.
+ </p>
+
+ </li>
+
+ <li>
+ Make a more accessible plugin architecture for lldb.
+
+ <p>
+ Right now, you can only use the Python or SB API's to extend an extant lldb. You can't
+ implement any of the actual lldb Plugins as plugins. That means anybody that wants to
+ add new Object file/Process/Language etc support has to build and distribute their own
+ lldb. This is tricky because the API's the plugins use are currently not stable (and
+ recently have been changing quite a lot.) We would have to define a subset of lldb_private
+ that you could use, and some way of telling whether the plugins were compatible with
+ the lldb. But long-term, making this sort of extension possible will make lldb more
+ appealing for research and 3rd party uses.
+ </p>
+ </li>
+
+ <li>
+ Use instruction emulation to avoid the overhead of swapping trap and instruction for breakpoints
+
+ <p>
+ At present, breakpoints are implemented by inserting a trap instruction, then when the trap
+ is hit, replace the trap with the actual instruction and single step. Then swap back and
+ continue. This causes problems for read only text, and also means that no-stop debugging
+ ust either stop all threads briefly to handle this two-step or risk missing some
+ breakpoint hits. If you emulated the instruction and wrote back the results, you wouldn't
+ have these problems, and it would also save a stop per breakpoint hit. Since we use breakpoints
+ to implement stepping, this savings could be significant on slow connections.
+ </p>
+ </li>
+ <li>
+ Use the JIT to speed up conditional breakpoint evaluation
+
+ <p>
+ We already JIT and cache the conditional expressions for breakpoints for the C family of
+ languages, so we aren't re-compiling every time you hit the breakpoint. And if we couldn't
+ IR interpret the expression, we leave the JIT'ed code in place for reuse. But it would
+ be even better if we could also insert the "stop or not" decision into the code at the
+ breakpoint, so you would only actually stop the process when the condition was true.
+ Greg's idea was that if you had a conditional breakpoint set when you started the
+ debug session, Xcode could rebuild and insert enough no-ops that we could instrument
+ the breakpoint site and call the conditional expression, and only trap if the conditional was true.
+ </p>
+ </li>
+
+ <li>
+ Broaden the idea in "target stop-hook" to cover more events in the debugger
+
+ <p>
+ Shared library loads, command execution, User directed memory/register reads and writes
+ are all places where you would reasonably want to hook into the debugger.
+ </p>
+ </li>
+
+ <li>
+ Mock classes for testing
+ <p>
+ We need "ProcessMock" and "ObjectFileMock" and the like. These would be real
+ plugin implementations for their underlying lldb classes, with the addition
+ that you can prime them from some sort of text based input files. For classes
+ that manage changes over time (like process) you would need to program the state
+ at StopPoint 0, StopPoint 1, etc. These could then be used for testing reactions
+ to complex threading
+ problems & the like, and also for simulating hard-to-test environments (like bare
+ board debugging).
+ </p>
+ </li>
+
+ <li>
+ A Bug-Trapper infrastructure
+ <p>
+ We very often have bugs that can't be reproduced locally. So having a bug-report-trapper
+ that can gather enough information from the
+ surroundings of a bug so that we can replay the session locally would be a big help
+ tracking down issues in this situation. This is tricky because you
+ can't necessarily require folks to leak information about their code in order to
+ file bug reports. So not only will you have to figure out what state to gather,
+ you're also going to have to anonymize it somehow. But we very often have bugs
+ from people that can't reduce the problem to a simple test case and can't give us
+ our code, and we often just can't help them as things stand now. Note that adding
+ the ProcessMock would be a good first stage towards this, since you could make a ProcessMock
+ creator/serializer from the current lldb state.
+ </p>
+ </li>
+
+ <li>
+ Expression parser needs syntax for "{symbol,type} A in CU B.cpp" etc.
+
+ <p>
+ Sometimes you need to specify non-visible or ambiguous types to the expression parser.
+ We were planning to do $b_dot_cpp$A or something like. You might want to specify a
+ static in a function, in a source file, or in a shared library. So the syntax should
+ support all these.
+ </p>
+ </li>
+
+ <li>
+ Add a "testButDontAbort" style test to the UnitTest framework.
+
+ <p>
+ The way we use unittest now (maybe this is the only way it can work, I don't know)
+ you can't report a real failure and continue with the test. That is appropriate
+ in some cases: if I'm supposed to hit breakpoint A before I evaluate an expression,
+ and don't hit breakpoint A, the test should fail. But it means that if I want to
+ test five different expressions, I can either do it in one test, which is good
+ because it means I only have to fire up one process, attach to it, and get it to
+ a certain point. But it also means if the first test fails, the other four don't
+ even get run. So though at first we wrote a bunch of test like this, as time went
+ on we switched more to writing "one at a time" tests because they were more robust
+ against a single failure. That makes the test suite run much more slowly. It
+ would be great to add a "test_but_dont_abort" variant of the tests, then we could
+ gang tests that all drive to the same place and do similar things. As an added
+ benefit, it would allow us to be more thorough in writing tests, since each
+ test would have lower costs.
+ </p>
+ </li>
+
+ <li>
+ Unify Watchpoint's & Breakpoints.
+ <p>
+ Option handling isn't shared, and more importantly the PerformAction's have a lot
+ of duplicated common code, most of which works less well on the Watchpoint side.
+ </p>
+ </li>
+
+ <li>
+ Reverse debugging.
+ <p>
+ This is kind of a holy grail, it's hard to support for complex apps (many
+ threads, shared memory, etc.) But it would be SO nice to have...
+ </p>
+ </li>
+
+ <li>
+ Non-stop debugging.
+
+ <p>
+ By this I mean allowing some threads in the target program to run while stopping
+ other threads. This is supported in name in lldb at present, but lldb makes the
+ assumption "If I get a stop, I won't get another stop unless I actually run the
+ program." in a bunch of places so getting it to work reliably will be some a
+ good bit of work. And figuring out how to present this in the UI will also be tricky.
+ </p>
+ </li>
+
+ <li>
+ Fix and continue.
+ <p>
+ We did this in gdb without a real JIT. The implementation shouldn't be that hard,
+ especially if you can build the executable for fix and continue. The tricky part is
+ how to verify that the user can only do the kinds of fixes that are safe to do.
+ No changing object sizes is easy to detect, but there were many more subtle changes
+ (function you are fixing is on the stack...) that take more work to prevent.
+ And then you have to explain these conditions the user in some helpful way.
+ </p>
+ </li>
+ </ol>
+ </div>
+ <div class="postfooter"></div>
+ </div>
+ </div>
+ </div>
+</div>
+</body>
+</html>
Propchange: lldb/trunk/www/projects.html
------------------------------------------------------------------------------
svn:executable = *
Modified: lldb/trunk/www/sidebar.incl
URL: http://llvm.org/viewvc/llvm-project/lldb/trunk/www/sidebar.incl?rev=296814&r1=296813&r2=296814&view=diff
==============================================================================
--- lldb/trunk/www/sidebar.incl (original)
+++ lldb/trunk/www/sidebar.incl Thu Mar 2 15:39:27 2017
@@ -8,6 +8,7 @@
<li><a href="/goals.html">Goals</a></li>
<li><a href="/features.html">Features</a></li>
<li><a href="/status.html">Status</a></li>
+ <li><a href="/projects.html">Projects</a></li>
</ul>
</div>
<div class="menu">
More information about the lldb-commits
mailing list