[Openswan dev] [PATCH] silent wrong uninitialized compiler warning

[D. Hugh Redelmeier]

| From: Gilles Espinasse <g.esp at free.fr>

| From: "D. Hugh Redelmeier" <hugh at mimosa.com>
| To: "Gilles Espinasse" <g.esp at free.fr>; <dev at openswan.org>
| Sent: Wednesday, June 24, 2009 8:27 AM
| Subject: Re: [Openswan dev] [PATCH] silent wrong uninitialized compiler
| warning
| 
| 
| > | From: Gilles Espinasse <g.esp at free.fr>

| > |  int nat_traversal_espinudp_socket (int sk, const char *fam, u_int32_t
| type)
| > |  {
| > | - int r;
| > | + int r = 0;

| > I took a quick look at the code in 2.6.22.  I don't understand its
| > complexity.  It seems as if the test in the first if statement has to
| > always be true since style is initialized to auto_style.  So lots of
| > the code is redundant an r is surely set before it is used.

| style is declared static.

| so there is no dead code there.

Oops!  Right you are.


| > It looks as if -1 is a better value to initialize r to, if you have to
| > do so.
| >
| I don't understand the reason.

-1 means "error".  Surely using the default value is an error.

And still a good idea to add a comment:
   /* initializing to shut up GCC warning about uninitialized use */


| > | diff --git a/programs/pluto/rnd.c b/programs/pluto/rnd.c

| > |      unsigned int max_rnd_devices = elemsof(random_devices)+1;

| > The code you are fixing here is overly complex too.  While trying to
| > understand it, I found a bug: off-by-one in the for loop.  The
| > definition of max_rnd_devices should not have the "+1".

| I have not written that code!
| That's true the +1 is wrong there.
| In the case all devices fail to open, openswan_log will access
| random_devices[max_rnd_devices].

Glad I got that right.