[PATCH] bloat-o-meter: teach it to handle aliases

[Bernhard Reutner-Fischer]

On Sun, Feb 07, 2010 at 11:40:33AM +0100, Bernhard Reutner-Fischer wrote:
>On Sat, Feb 06, 2010 at 11:15:42PM +0100, Denys Vlasenko wrote:
>>Hi Bernhard,
>
>>> I did not check the patch per se. The fix is conceptually correct.
>>> Please apply.
>>
>>Now, when it is in git and I actually used it...
>>make bloatcheck takes about 2 seconds on 2.6 GHz machine...
>>is this expected?
>
>it takes 0m1.158s with cold caches (and 0m0.856s with hot cache) on my 2.4GHz
>machine which is slightly slower than the old nm version.
>
>bloat-o-meter -t busybox_unstripped busybox_unstripped ¹) gives for me:
>374343168/383921920; 73914880 Parse origin/new; processing nsecs

fixed quadric behaviour in git.

$ time ./scripts/bloat-o-meter -t _bb_un.orig _bb_un.00 
function                                             old     new   delta
flashcp_main                                           -     835    +835
.rodata                                           135072  135537    +465
packed_usage                                       26028   26128    +100
progress                                               -      67     +67
applet_names                                        2102    2110      +8
applet_main                                         2496    2504      +8
applet_nameofs                                       624     626      +2
applet_install_loc                                   156     157      +1
------------------------------------------------------------------------------
(add/remove: 3/1 grow/shrink: 6/0 up/down: 1486/0)           Total: 1486
bytes

52937984/51841024; 9788160 Parse origin/new; processing nsecs
total nsecs: 114578944

real	0m0.135s
user	0m0.108s
sys	0m0.024s

before the alias resolution the same two binaries took:
real	0m0.073s
user	0m0.048s
sys	0m0.036s

So it takes about twice the time for the alias resolution, which is
expected.