|
1 New Trace Compiler |
|
2 ------------------- |
|
3 |
|
4 Raptor supports an updated version of the Trace Compiler which has |
|
5 an improved command-line interface that allows Raptor to specify the |
|
6 directory in which trace data should be written. This allows MMPs to |
|
7 implement a new convention for where trace files are stored. |
|
8 |
|
9 The new convention prevents clashes where building the same |
|
10 source file for different MMPs would previously have resulted in two |
|
11 build tasks trying to create the same trace file (header/dictionary). |
|
12 This caused build slowdowns and also made the trace data inaccurate. |
|
13 |
|
14 The standard for this directory is that the last two elements of the |
|
15 path should be |
|
16 |
|
17 traces/traces_<targetname>_<ext> |
|
18 |
|
19 e.g. |
|
20 |
|
21 traces/traces_euser_dll |
|
22 |
|
23 The new trace compiler also creates "autogen" headers in a new location |
|
24 under epoc32/include/platform rather than epoc32/include/internal. |
|
25 |
|
26 Trace Headers Directory can now also be in SYSTEMINCLUDEs |
|
27 ---------------------------------------------------------- |
|
28 |
|
29 Some teams were having trouble getting qmake to generate MMPs with |
|
30 USERINCLUDEs set. So they were specifying the trace header directory in |
|
31 their SYTEMINCLUDEs. This change causes Raptor to look for the traces |
|
32 directory in bot SYSTEMINCLUDES and USERINCLUDES. |
|
33 |
|
34 |
|
35 Create Trace Directories During Makefile Parse |
|
36 ----------------------------------------------- |
|
37 |
|
38 Some build engines consider it a "clash" when two jobs attempt to make the |
|
39 same directory. This change causes some important trace output directories |
|
40 to be created during makefile parsing which removes the (small) possibility |
|
41 that they would clash during the build part. |
|
42 |