1On Demand Symbols 2================= 3 4On demand symbols can be enabled in LLDB for projects that generate debug info 5for more than what is required by a normal debug session. Some build systems 6enable debug information for all binaries and can end up producing many 7gigabytes of debug information. This amount of debug information can greatly 8increase debug session load times and can slow developer productivity when the 9debug information isn't indexed. It can also cause expression evaluation to 10be slow when types from all of the binaries have full debug info as each module 11is queried for very common types, or global name lookups fail due to a mistyped 12expression. 13 14When should I consider enabling this feature? 15--------------------------------------------- 16 17Anyone that has a build system that produces debug information for many 18binaries that are not all required when you want to focus on debugging a few of 19the produced binaries. Some build systems enable debug info as a project wide 20switch and the build system files that control how things are built are not 21easy to modify to produce debug info for only a small subset of the files being 22linked. If your debug session startup times are slow because of too much debug 23info, this feature might help you be more productive during daily use. 24 25How do I enable on demand symbols? 26---------------------------------- 27 28This feature is enabled using a LLDB setting: 29 30 31:: 32 33 (lldb) settings set symbols.load-on-demand true 34 35Users can also put this command into their ~/.lldbinit file so it is always 36enabled. 37 38How does this feature work? 39--------------------------- 40 41This feature works by selectively enabling debug information for modules that 42the user focuses on. It is designed to be enabled and work without the user 43having to set any other settings and will try and determine when to enable 44debug info access from the modules automatically. All modules with debug info 45start off with their debug information turned off for expensive name and type 46lookups. The debug information for line tables are always left enabled to allow 47users to reliably set breakpoints by file and line. As the user debugs their 48target, some simple things can cause module to get its debug information 49enabled (called hydration): 50- Setting a file and line breakpoint 51- Any PC from any stack frame that maps to a module 52- Setting a breakpoint by function name 53- Finding a global variable by name 54 55Since most users set breakpoints by file and line, this is an easy way for 56people to inform the debugger that they want focus on this module. Breakpoints 57by file and line are always enabled when on demand symbols is being used. Line 58tables in debug information are cheap to parse and breakpoints will be able to 59be set in any module that has debug info. Setting breakpoints by file and line 60acts as one of the primary ways to enable debug info for a module as it is 61the most common way to stop your program at interesting areas of your code. 62 63Once the user hits a breakpoint, or stops the program for any other reason, 64like a crash, assertion or signal, the debugger will calculate the stack frames 65for one or more threads. Any stack frames whose PC value is contained within 66one of a module's sections will have its debug information enabled. This allows 67us to enable debug information for the areas of code that the user is stopped 68in and will allow only the important subset of modules to have their debug 69information enabled. 70 71On demand symbol loading tries to avoid indexing the names within the debug 72information and makes a few tradeoffs to allow name matching of functions and 73globals without having to always index all of the debug information. 74Setting breakpoints by function name can work, but we try to avoid using 75debug information by relying on the symbol tables from a module. Debug 76information name indexing is one of the most expensive operations that we are 77trying to avoid with the on demand symbol loading so this is one of the main 78tradeoffs of this feature. When setting a breakpoint by function name, if the 79symbol table contains a match, the debug information will be enabled for that 80module and the query will be completed using the debug information. This does 81mean that setting breakpoints on inline function names can fail for modules 82that have debug info, but no matches in the symbol table since inlined 83functions don't exist in symbol tables. When using on demand symbol loading it 84is encouraged to not strip the symbol tables of local symbols as this will 85allow users to set breakpoints on all concrete functions reliably. Stripped 86symbol tables have their local symbols removed from the symbol table which 87means that static functions and non exported function will not appear in the 88symbol tables. This can cause breakpoint setting by function name to fail when 89previously it wouldn't fail. 90 91Global variable lookups rely on the same methodology as breakpoint setting by 92function name: we use the symbol tables to find a match first if debug 93information isn't enabled for a module. If we find a match in the symbol table 94for a global variable lookup, we will enable debug information and complete 95the query using the debug information. It is encouraged to not strip your 96symbol tables with this features as static variables and other non exported 97globals will not appear in the symbol table and can lead to matches not being 98found. 99 100What other things might fail? 101----------------------------- 102 103The on demand symbol loading feature tries to limit expensive name lookups 104within debug information. As such, some lookups by name might fail when they 105wouldn't when this feature is not enabled: 106- Setting breakpoints by function name for inlined functions 107- Type lookups when the expression parser requests types by name 108- Global variable lookups by name when the name of the variable is stripped 109 110Setting breakpoints by function name can fail for inline function because this 111information is only contained in the debug information. No symbols are created 112for inlined functions unless there is a concrete copy of the inlined function 113in that same module. As a result, we might not end up stopping at all inlined 114functions when requested with this feature enabled. Setting file and line 115breakpoints are a good way still use on demand symbol loading effectively 116and still being able to stop at inline function invocations. 117 118The expression parser often tries to lookup types by name when the user types 119an expression. These are some of the most costly parts of expression evaluation 120as the user can type something like "iterator" as part of their expression and 121this can result in matches from all STL types in all modules. These kinds of 122global type lookup queries can cause thousands of results to be found if debug 123information is enabled. The way that most debug information is created these 124days has the type information inlined into each module. Typically each module 125will contain full type definitions in the debug info for any types that are 126used in code. This means that when you type an expression when stopped, you 127have debug information for all of the variables, arguments and global variables 128in your current stack frame and we should be able to find type that are 129important by using only the modules that have their debug information enabled. 130 131The expression parser can also be asked to display global variables and they 132can be looked up by name. For this feature to work reliably with on demand 133symbol loading enabled, just don't strip your symbol tables and the expression 134parser should have no problem finding your variables. Most global variables 135that are exported will still be in your symbol table if it is stripped, but 136static variables and non exported globals will not be. 137 138Can I troubleshoot issues when I believe this feature is impeding my debugging? 139------------------------------------------------------------------------------- 140 141Logging has been added that can be enabled to help notify our engineers when 142something is not producing results when this feature is enabled. This logging 143can be enabled during a debug session and can be sent to the LLDB engineers 144to help troubleshoot these situation. To enable logging, type the following 145command: 146 147:: 148 149 (lldb) log enable -f /tmp/ondemand.txt lldb on-demand 150 151When the logging is enabled, we get full visibility into each query that would 152have produced results if this feature were not enabled and will allow us to 153troublshoot issues. Enabling this logging before an expression, setting a 154breakpoint by name, or doing a type lookup can help us see the patterns that 155cause failures and will help us improve this feature. 156