1Open Projects 2============= 3 4The following is a mostly unordered set of the ideas for improvements to the 5LLDB debugger. Some are fairly deep, some would require less effort. 6 7Speed up type realization in lldb 8--------------------------------- 9 10The type of problem I'm addressing here is the situation where you are 11debugging a large program (lldb built with debug clang/swift will do) and you 12go to print a simple expression, and lldb goes away for 30 seconds. When you 13sample it, it is always busily churning through all the CU's in the world 14looking for something. The problem isn't that looking for something in 15particular is slow, but rather that we somehow turned an bounded search (maybe 16a subtype of "std::string" into an unbounded search (all things with the name 17of that subtype.) Or didn't stop when we got a reasonable answer proximate to 18the context of the search, but let the search leak out globally. And quite 19likely there are other issues that I haven't guessed yet. But if you end up 20churning though 3 or 4 Gig of debug info, that's going to be slow no matter how 21well written your debug reader is... 22 23My guess is the work will be more in the general symbol lookup than in the 24DWARF parser in particular, but it may be a combination of both. 25 26As a user debugging a largish program, this is the most obvious lameness of 27lldb. 28 29Symbol name completion in the expression parser 30----------------------------------------------- 31 32This is the other obvious lameness of lldb. You can do: 33 34:: 35 36 (lldb) frame var foo.b 37 38and we will tell you it is "foo.bar". But you can't do that in the expression 39parser. This will require collaboration with the clang/swift folks to get the 40right extension points in the compiler. And whatever they are, lldb will need 41use them to tell the compiler about what names are available. It will be 42important to avoid the pitfalls of #1 where we wander into the entire DWARF 43world. 44 45Make a high speed asynchronous communication channel 46---------------------------------------------------- 47 48All lldb debugging nowadays is done by talking to a debug agent. We used the 49gdb-remote protocol because that is universal, and good enough, and you have to 50support it anyway since so many little devices & JTAG's and VM's etc support 51it. But it is really old, not terribly high performance, and can't really 52handle sending or receiving messages while the process is supposedly running. 53It should have compression built in, remove the hand-built checksums and rely 54on the robust communication protocols we always have nowadays, allow for 55out-of-order requests/replies, allow for reconnecting to a temporarily 56disconnected debug session, regularize all of the packet formatting into JSON 57or BSON or whatever while including a way to do large binary transfers. It must 58be possible to come up with something faster, and better tunable for the many 59communications pathways we end up supporting. 60 61Fix local variable lookup in the lldb expression parser 62------------------------------------------------------- 63 64The injection of local variables into the clang expression parser is 65currently done incorrectly - it happens too late in the lookup. This results 66in namespace variables & functions, same named types and ivars shadowing 67locals when it should be the other way around. An attempt was made to fix 68this by manually inserting all the visible local variables into wrapper 69function in the expression text. This mostly gets the job done but that 70method means you have to realize all the types and locations of all local 71variables for even the simplest of expressions, and when run on large 72programs (e.g. lldb) it would cause unacceptable delays. And it was very 73fragile since an error in realizing any of the locals would cause all 74expressions run in that context to fail. We need to fix this by adjusting 75the points where name lookup calls out to lldb in clang. 76 77Support calling SB & commands everywhere and support non-stop debugging 78----------------------------------------------------------------------- 79 80There is a fairly ad-hoc system to handle when it is safe to run SB API's and 81command line commands. This is actually a bit of a tricky problem, since we 82allow access to the command line and SB API from some funky places in lldb. The 83Operating System plugins are the most obvious instance, since they get run 84right after lldb is told by debugserver that the process has stopped, but 85before it has finished collating the information from the stop for presentation 86to the higher levels. But breakpoint callbacks have some of the same problems, 87and other things like the scripted stepping operations and any fancier 88extension points we want to add to the debugger are going to be hard to 89implement robustly till we work on a finer-grained and more explicit control 90over who gets to control the process state. 91 92We also won't have any chance of supporting non-stop debugging - which is a 93useful mode for programs that have a lot of high-priority or real-time worker 94threads - until we get this sorted out. 95 96Finish the language abstraction and remove all the unnecessary API's 97-------------------------------------------------------------------- 98 99An important part of making lldb a more useful "debugger toolkit" as opposed to 100a C/C++/ObjC/Swift debugger is to have a clean abstraction for language 101support. We did most, but not all, of the physical separation. We need to 102finish that. And then by force of necessity the API's really look like the 103interface to a C++ type system with a few swift bits added on. How you would 104go about adding a new language is unclear and much more trouble than it is 105worth at present. But if we made this nice, we could add a lot of value to 106other language projects. 107 108Add some syntax to generate data formatters from type definitions 109----------------------------------------------------------------- 110 111Uses of the data formatters fall into two types. There are data formatters for 112types where the structure elements pretty much tell you how to present the 113data, you just need a little expression language to express how to turn them 114into what the user expects to see. Then there are the ones (like pretty much 115all our Foundation/AppKit/UIKit formatters) that use deep magic to figure out 116how the type is actually laid out. The latter are pretty much always going to 117have to be done by hand. 118 119But for the ones where the information is expressed in the fields, it would be 120great to have a way to express the instructions to produce summaries and 121children in some form you could embed next to the types and have the compiler 122produce a byte code form of the instructions and then make that available to 123lldb along with the library. This isn't as simple as having clang run over the 124headers and produce something from the types directly. After all, clang has no 125way of knowing that the interesting thing about a std::vector is the elements 126that you get by calling size (for the summary) and [] for the elements. But it 127shouldn't be hard to come up with a generic markup to express this. 128 129Allow the expression parser to access dynamic type/data formatter information 130----------------------------------------------------------------------------- 131 132This seems like a smaller one. The symptom is your object is Foo child of 133Bar, and in the Locals view you see all the fields of Foo, but because the 134static type of the object is Bar, you can't see any of the fields of Foo. 135But if you could get this working, you could hijack the mechanism to make 136the results of the value object summaries/synthetic children available to 137expressions. And if you can do that, you could add other properties to an 138object externally (through Python or some other extension point) and then 139have these also available in the expression parser. You could use this to 140express invariants for data structures, or other more advanced uses of types 141in the debugger. 142 143Another version of this is to allow access to synthetic children in the 144expression parser. Otherwise you end up in situations like: 145 146:: 147 148 (lldb) print return_a_foo() 149 (SomeVectorLikeType) $1 = { 150 [0] = 0 151 [1] = 1 152 [2] = 2 153 [3] = 3 154 [4] = 4 155 } 156 157That's good but: 158 159:: 160 161 (lldb) print return_a_foo()[2] 162 163fails because the expression parser doesn't know anything about the 164array-like nature of SomeVectorLikeType that it gets from the synthetic 165children. 166 167Recover thread information lazily 168--------------------------------- 169 170LLDB stores all the user intentions for a thread in the ThreadPlans stored in 171the Thread class. That allows us to reliably implement a very natural model for 172users moving through a debug session. For example, if step-over stops at a 173breakpoint in an function in a younger region of the stack, continue will 174complete the step-over rather than having to manually step out. But that means 175that it is important that the Thread objects live as long as the Threads they 176represent. For programs with many threads, but only one that you are debugging, 177that makes stepping less efficient, since now you have to fetch the thread list 178on every step or stepping doesn't work correctly. This is especially an issue 179when the threads are provided by an Operating System plugin, where it may take 180non-trivial work to reconstruct the thread list. It would be better to fetch 181threads lazily but keep "unseen" threads in a holding area, and only retire 182them when we know we've fetched the whole thread list and ensured they are no 183longer alive. 184 185Make Python-backed commands first class citizens 186------------------------------------------------ 187 188As it stands, Python commands have no way to advertise their options. They are 189required to parse their arguments by hand. That leads to inconsistency, and 190more importantly means they can't take advantage of auto-generated help and 191command completion. This leaves python-backed commands feeling worse than 192built-in ones. 193 194As part of this job, it would also be great to hook automatically hook the 195"type" of an option value or argument (e.g. eArgTypeShlibName) to sensible 196default completers. You need to be able to over-ride this in more complicated 197scenarios (like in "break set" where the presence of a "-s" option limits the 198search for completion of a "-n" option.) But in common cases it is unnecessary 199busy-work to have to supply the completer AND the type. If this worked, then it 200would be easier for Python commands to also get correct completers. 201 202Reimplement the command interpreter commands using the SB API 203------------------------------------------------------------- 204 205Currently, all the CommandObject::DoExecute methods are implemented using the 206lldb_private API's. That generally means that there's code that gets duplicated 207between the CommandObject and the SB API that does roughly the same thing. We 208would reduce this code duplication, present a single coherent face to the users 209of lldb, and keep ourselves more honest about what we need in the SB API's if 210we implemented the CommandObjects::DoExecute methods using the SB API's. 211 212BTW, it is only the way it was much easier to develop lldb if it had a 213functioning command-line early on. So we did that first, and developed the SB 214API's when lldb was more mature. There's no good technical reason to have the 215commands use the lldb_private API's. 216 217Documentation and better examples 218--------------------------------- 219 220We need to put the lldb syntax docs in the tutorial somewhere that is more 221easily accessible. On suggestion is to add non-command based help to the help 222system, and then have a "help lldb" or "help syntax" type command with this 223info. Be nice if the non-command based help could be hierarchical so you could 224make topics. 225 226There's a fair bit of docs about the SB API's, but it is spotty. Some classes 227are well documented in the Python "help (lldb.SBWhatever)" and some are not. 228 229We need more conceptual docs. And we need more examples. And we could provide a 230clean pluggable example for using LLDB standalone from Python. The 231process_events.py is a start of this, but it just handles process events, and 232it is really a quick sketch not a polished expandable proto-tool. 233 234Make a more accessible plugin architecture for lldb 235--------------------------------------------------- 236 237Right now, you can only use the Python or SB API's to extend an extant lldb. 238You can't implement any of the actual lldb Plugins as plugins. That means 239anybody that wants to add new Object file/Process/Language etc support has to 240build and distribute their own lldb. This is tricky because the API's the 241plugins use are currently not stable (and recently have been changing quite a 242lot.) We would have to define a subset of lldb_private that you could use, and 243some way of telling whether the plugins were compatible with the lldb. But 244long-term, making this sort of extension possible will make lldb more appealing 245for research and 3rd party uses. 246 247Use instruction emulation to reduce the overhead for breakpoints 248---------------------------------------------------------------- 249 250At present, breakpoints are implemented by inserting a trap instruction, then 251when the trap is hit, replace the trap with the actual instruction and single 252step. Then swap back and continue. This causes problems for read only text, and 253also means that no-stop debugging must either stop all threads briefly to handle 254this two-step or risk missing some breakpoint hits. If you emulated the 255instruction and wrote back the results, you wouldn't have these problems, and 256it would also save a stop per breakpoint hit. Since we use breakpoints to 257implement stepping, this savings could be significant on slow connections. 258 259Use the JIT to speed up conditional breakpoint evaluation 260--------------------------------------------------------- 261 262We already JIT and cache the conditional expressions for breakpoints for the C 263family of languages, so we aren't re-compiling every time you hit the 264breakpoint. And if we couldn't IR interpret the expression, we leave the JIT'ed 265code in place for reuse. But it would be even better if we could also insert 266the "stop or not" decision into the code at the breakpoint, so you would only 267actually stop the process when the condition was true. Greg's idea was that if 268you had a conditional breakpoint set when you started the debug session, Xcode 269could rebuild and insert enough no-ops that we could instrument the breakpoint 270site and call the conditional expression, and only trap if the conditional was 271true. 272 273Broaden the idea in "target stop-hook" to cover more events in the debugger 274--------------------------------------------------------------------------- 275 276Shared library loads, command execution, User directed memory/register reads 277and writes are all places where you would reasonably want to hook into the 278debugger. 279 280Mock classes for testing 281------------------------ 282 283We need "ProcessMock" and "ObjectFileMock" and the like. These would be real 284plugin implementations for their underlying lldb classes, with the addition 285that you can prime them from some sort of text based input files. For classes 286that manage changes over time (like process) you would need to program the 287state at StopPoint 0, StopPoint 1, etc. These could then be used for testing 288reactions to complex threading problems & the like, and also for simulating 289hard-to-test environments (like bare board debugging). 290 291Expression parser needs syntax for "{symbol,type} A in CU B.cpp" 292---------------------------------------------------------------- 293 294Sometimes you need to specify non-visible or ambiguous types to the expression 295parser. We were planning to do $b_dot_cpp$A or something like. You might want 296to specify a static in a function, in a source file, or in a shared library. So 297the syntax should support all these. 298 299Add a "testButDontAbort" style test to the UnitTest framework 300------------------------------------------------------------- 301 302The way we use unittest now (maybe this is the only way it can work, I don't 303know) you can't report a real failure and continue with the test. That is 304appropriate in some cases: if I'm supposed to hit breakpoint A before I 305evaluate an expression, and don't hit breakpoint A, the test should fail. But 306it means that if I want to test five different expressions, I can either do it 307in one test, which is good because it means I only have to fire up one process, 308attach to it, and get it to a certain point. But it also means if the first 309test fails, the other four don't even get run. So though at first we wrote a 310bunch of test like this, as time went on we switched more to writing "one at a 311time" tests because they were more robust against a single failure. That makes 312the test suite run much more slowly. It would be great to add a 313"test_but_dont_abort" variant of the tests, then we could gang tests that all 314drive to the same place and do similar things. As an added benefit, it would 315allow us to be more thorough in writing tests, since each test would have lower 316costs. 317 318Convert the dotest style tests to use lldbutil.run_to_source_breakpoint 319----------------------------------------------------------------------- 320 321run_to_source_breakpoint & run_to_name_breakpoint provide a compact API that 322does in one line what the first 10 or 20 lines of most of the old tests now do 323by hand. Using these functions makes tests much more readable, and by 324centralizing common functionality will make maintaining the testsuites easier 325in the future. This is more of a finger exercise, and perhaps best implemented 326by a rule like: "If you touch a test case, and it isn't using 327run_to_source_breakpoint, please make it do so". 328 329Unify Watchpoint's & Breakpoints 330-------------------------------- 331 332Option handling isn't shared, and more importantly the PerformAction's have a 333lot of duplicated common code, most of which works less well on the Watchpoint 334side. 335 336Reverse debugging 337----------------- 338 339This is kind of a holy grail, it's hard to support for complex apps (many 340threads, shared memory, etc.) But it would be SO nice to have... 341 342Non-stop debugging 343------------------ 344 345By this I mean allowing some threads in the target program to run while 346stopping other threads. This is supported in name in lldb at present, but lldb 347makes the assumption "If I get a stop, I won't get another stop unless I 348actually run the program." in a bunch of places so getting it to work reliably 349will be some a good bit of work. And figuring out how to present this in the UI 350will also be tricky. 351 352Fix and continue 353---------------- 354 355We did this in gdb without a real JIT. The implementation shouldn't be that 356hard, especially if you can build the executable for fix and continue. The 357tricky part is how to verify that the user can only do the kinds of fixes that 358are safe to do. No changing object sizes is easy to detect, but there were many 359more subtle changes (function you are fixing is on the stack...) that take more 360work to prevent. And then you have to explain these conditions the user in some 361helpful way. 362 363Unified IR interpreter 364---------------------- 365 366Currently IRInterpreter implements a portion of the LLVM IR, but it doesn't 367handle vector data types and there are plenty of instructions it also doesn't 368support. Conversely, lli supports most of LLVM's IR but it doesn't handle 369remote memory and its function calling support is very rudimentary. It would be 370useful to unify these and make the IR interpreter -- both for LLVM and LLDB -- 371better. An alternate strategy would be simply to JIT into the current process 372but have callbacks for non-stack memory access. 373 374Teach lldb to predict exception propagation at the throw site 375------------------------------------------------------------- 376 377There are a bunch of places in lldb where we need to know at the point where an 378exception is thrown, what frame will catch the exception. 379 380For instance, if an expression throws an exception, we need to know whether the 381exception will be caught in the course of the expression evaluation. If so it 382would be safe to let the expression continue. But since we would destroy the 383state of the thread if we let the exception escape the expression, we currently 384stop the expression evaluation if we see a throw. If we knew where it would be 385caught we could distinguish these two cases. 386 387Similarly, when you step over a call that throws, you want to stop at the throw 388point if you know the exception will unwind past the frame you were stepping in, 389but it would annoying to have the step abort every time an exception was thrown. 390If we could predict the catching frame, we could do this right. 391 392And of course, this would be a useful piece of information to display when stopped 393at a throw point. 394 395Add predicates to the nodes of settings 396--------------------------------------- 397 398It would be very useful to be able to give values to settings that are dependent 399on the triple, or executable name, for targets, or on whether a process is local 400or remote, or on the name of a thread, etc. The original intent (and there is 401a sketch of this in the settings parsing code) was to be able to say: 402 403:: 404 405 (lldb) settings set target{arch=x86_64}.process.thread{name=foo}... 406 407The exact details are still to be worked out, however. 408 409Resurrect Type Validators 410------------------------- 411 412This half-implemented feature was removed in 413https://reviews.llvm.org/D71310 but the general idea might still be 414useful: Type Validators look at a ValueObject, and make sure that 415there is nothing semantically wrong with the object's contents to 416easily catch corrupted data. 417