seed random with a real random data #377
Conversation
| @@ -0,0 +1,59 @@ | |||
| #ifdef WIN32 | |||
| #include "Windows.h" | |||
|
I really want to start work on fixing the why we use random and other bits ;) This is the start. @ossec anyone able to review? |
| #endif | ||
|
|
||
| if (failed) { | ||
| ErrorExit("Error in randombytes failed on all possiable methods for accessing random data"); |
|
typo fixed. Building now. |
| #endif | ||
|
|
||
| if (failed) { | ||
| ErrorExit("Error in randombytes failed on all possible methods for accessing random data"); |
There was a problem hiding this comment.
I agree with @cgzones suggestion here:
Having some more context around what application had the error using __local_name would be useful here I think.
| #endif | ||
|
|
||
| if (failed) { | ||
| ErrorExit("%s: Error in randombytes failed on all possible methods for accessing random data", __local_name); |
There was a problem hiding this comment.
Better but can we make it the same formatting many of the other error messages get. Here is a random example I found:
https://github.com/ossec/ossec-hids/blob/master/src/shared/file_op.c#L454
ErrorExit("%s: ERROR: Message goes here", __local_name);
Wish this was abstracted a bit better so you didn't have to do this each time.
|
Was just about to merge this but looks like you have to merge master before I can. |
|
@awiddersheim fixed the conflicts :) travis will be done in a few minutes. |
seed random with a real random data
The way random was seed was predictable. This does not fix it, but makes it better and standardizes the process across domains. I will be looking to use random bytes more often in the code base.
Randombytes started life: https://github.com/ultramancool/tweetnacl-usable/blob/master/randombytes.c thank you @ultramancool and tweetnacl/djb guys.