gethostbyname call with 0 length arg

[Phil Goss]

Hi,

Is it a bug in the WINE gethostbyname (WS_gethostbyname) implementation,  
where if the arg name is of string length 0, then it returns unsuccessful.

I have experience this issue with a Windows program (Onkyo Nettune) where  
the hostname passed to gethostbyname is of length 0. (This seems to be a  
bug in the program itself). However, the program works okay in native  
Windows.

There seems to be a check in the gethostbyname for a NULL arg name  ...  
if(!name) ... But no check for  ... if(!strlen(name)) ....

I have tested the strlen check and it seems to work and satisfies the  
Windows program.

Does the above make sense, as I am not a programmer?
Not sure if this s bug in wine even though it implementation should be  
matching Windows native.

Thanks,
Phil

[James Hawkins]

On 1/29/06, Phil Goss  wrote:
> Hi,
>
> Is it a bug in the WINE gethostbyname (WS_gethostbyname) implementation,
> where if the arg name is of string length 0, then it returns unsuccessful.

...
> I have tested the strlen check and it seems to work and satisfies the
> Windows program.

This would make a good regression test.  See:

http://winehq.org/site/docs/winedev-guide/testing

--
James Hawkins

[Detlef Riekenberg]

Am Sonntag, den 29.01.2006, 14:46 +0100 schrieb Phil Goss:
> There seems to be a check in the gethostbyname for a NULL arg name  ...  
> if(!name) ... But no check for  ... if(!strlen(name)) ....

Alexandre changed "strlen(xxx) > 0" in my first Patch to xxx[0]
So for your code above, simple change:

if(!name) {

to

if((!name) && (name[0])) {

The logic is the same: use the code-path, if name is NULL and use it
also, when the string is empty.

I send this type of patch yesterday and it's already in the tree.

The Commit-Message is here:
http://www.winehq.org/pipermail/wine-cvs/2006-February/020657.html

(i used (pName != NULL) to have a common style in this function)
> Does the above make sense, as I am not a programmer?

Yes, it does, and it's important when we have a Program that depend on
this behaviour.
> Not sure if this s bug in wine even though it implementation should be  
> matching Windows native.

That's also a reason, why we have the Testsuite.
A specific Test changes "should match" to 
"does not match (yet)" or "matches all the cases we tested".

The winedev-guide is nice to read. (see msg from James)

[Christer Palm]

Detlef Riekenberg wrote:
> if((!name) && (name[0])) {

Do you really mean that?
"if name is 0, then dereference name"?

--
Christer Palm

[Phil Goss]

This is what seems to have worked for me

if(!name || !strlen(name)) {

On Fri, 03 Feb 2006 19:23:22 +0100, Christer Palm  wrote:
> Detlef Riekenberg wrote:
> > if((!name) && (name[0])) {
>
> Do you really mean that?
> "if name is 0, then dereference name"?
>
> --
> Christer Palm

[James Hawkins]

On 2/3/06, Phil Goss  wrote:
> This is what seems to have worked for me
>
> if(!name || !strlen(name)) {

The shortest way to do this is:

if (!name || !*name) {

--
James Hawkins

[Detlef Riekenberg]

Am Samstag, den 04.02.2006, 00:59 +0100 schrieb Phil Goss:
> This is what seems to have worked for me
>
> if(!name || !strlen(name)) {

strlen() is overkill here, because you do not need the real length of
the string.

So "!name[0]" (or "!*name") instead of "!strlen(name)" is more
efficient, when you want to check for an empty name.

[Detlef Riekenberg]

Am Freitag, den 03.02.2006, 19:23 +0100 schrieb Christer Palm:
> > if((!name) && (name[0])) {
>
> Do you really mean that?
> "if name is 0, then dereference name"?

No. the code is:

"if name is not NULL, then dereference name"

[James Hawkins]

On 2/8/06, Detlef Riekenberg  wrote:
> Am Freitag, den 03.02.2006, 19:23 +0100 schrieb Christer Palm:
> > > if((!name) && (name[0])) {
> >
> > Do you really mean that?
> > "if name is 0, then dereference name"?
>
> No. the code is:
>
> "if name is not NULL, then dereference name"

Consider the case when name is NULL:

!name = TRUE

if (TRUE && name[0])

So the only time you'll dereference name is when name is NULL.  It should read:

if (name && name[0])

--
James Hawkins

[Phil Goss]

Thanks,

So basically, this check is now taken care of in the latest patches,  
correct?

Good ... Ill take a look at the guide as James suggested.

Cheers,
Phil

On Wed, 01 Feb 2006 23:27:53 +0100, Detlef Riekenberg   
wrote:
> Am Sonntag, den 29.01.2006, 14:46 +0100 schrieb Phil Goss:
> > There seems to be a check in the gethostbyname for a NULL arg name  ...
> > if(!name) ... But no check for  ... if(!strlen(name)) ....
>
> Alexandre changed "strlen(xxx) > 0" in my first Patch to xxx[0]
> So for your code above, simple change:
>
> if(!name) {
>
> to
>
> if((!name) && (name[0])) {
>
> The logic is the same: use the code-path, if name is NULL and use it
> also, when the string is empty.
>
> I send this type of patch yesterday and it's already in the tree.
>
> The Commit-Message is here:
> http://www.winehq.org/pipermail/wine-cvs/2006-February/020657.html
>
> (i used (pName != NULL) to have a common style in this function)
> > Does the above make sense, as I am not a programmer?
>
> Yes, it does, and it's important when we have a Program that depend on
> this behaviour.
> > Not sure if this s bug in wine even though it implementation should be
> > matching Windows native.
>
> That's also a reason, why we have the Testsuite.
> A specific Test changes "should match" to
> "does not match (yet)" or "matches all the cases we tested".
>
> The winedev-guide is nice to read. (see msg from James)