[libcxx-commits] [PATCH] D138601: [libc++] Fix memory leaks when throwing inside std::vector constructors

[David Blaikie via Phabricator via libcxx-commits]

dblaikie added a comment.

I ran at least a quick check, just `clang -O0 -g -gsplit-dwarf` self-host, basically, and the differences seem small:

     FILE SIZE        VM SIZE                                                                                                                
  --------------  --------------                                                                                                              
  +0.2%  +811Ki  [ = ]       0    .debug_info.dwo                                                                                            
  +0.2%  +612Ki  [ = ]       0    .debug_str.dwo                                                                                            
  +0.1% +84.3Ki  [ = ]       0    .debug_str_offsets.dwo                                                                                     
  +0.2% +17.2Ki  [ = ]       0    .debug_abbrev.dwo                                                                                          
  +0.1% +11.2Ki  [ = ]       0    .debug_rnglists.dwo                                                                                        
  +0.2% +1.50Mi  [ = ]       0    TOTAL  

      FILE SIZE        VM SIZE                                                                                                                
   --------------  --------------                                                                                                              
   +0.3%  +860Ki  [ = ]       0    .gdb_index    
  ...
   +0.2% +70.8Ki  [ = ]       0    .debug_addr                                                                                                
   +0.3% +35.7Ki  [ = ]       0    .debug_rnglists                                                                                            
   +0.1% +1.53Ki  [ = ]       0    .debug_info                                                                                                
   +0.1% +1.45Ki  [ = ]       0    .debug_str                                                                                                 
   +0.0%    +142  [ = ]       0    .debug_loclists                                                                                            
   +0.0%     +92  [ = ]       0    .debug_str_offsets                                                                                         
   +0.0%     +89  [ = ]       0    .debug_line_str             

So, that sounds OK to me (our reporting threshold is 1% overall file size, I think - then we drcill into the per section sizes to see what caused the problems) - though results for different programs and in different build modes might swing the numbers. I reached out to @joanahalili and co on our release team to see if they could run the wider release testing, but not sure if they'll have time - it's still a bit of a painful/long/unreliable process to run & there's work underway to streamline it.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D138601/new/

https://reviews.llvm.org/D138601