i2c-tools: do not fail opening bus if /dev/i2c

In case /dev/i2c is not a directory, open_i2c_dev will fail opening /dev/i2c/<i2cbus> but only handles the ENOENT errno, while we should also handle ENOTDIR errno because this is the returned error code. The patch attached fixes the issue.

Hi Florian,

Quoted 2 lines · Florian Fainelli

In case /dev/i2c is not a directory, open_i2c_dev will fail opening /dev/i2c/<i2cbus> but only handles the ENOENT errno, while we should

also

Quoted 1 line

handle ENOTDIR errno because this is the returned error code. The

patch

Quoted 1 line

attached fixes the issue.

This is better discussed on the linux-i2c list. Redirecting...

Quoted 5 lines

Signed-off-by Florian Fainelli <[email protected]>

--
Index: tools/i2cbusses.c
===================================================================
--- tools/.svn/text-base/i2cbusses.c.svn-base 2010-08-18

14:22:18.000000000 +0200

Quoted 1 line

+++ /tmp/tempfile.2.tmp 2010-08-18 14:33:12.000000000 +0200

What a ugly header. How am I supposed to apply this?

Quoted 4 lines

(quoted diff) +0 -0

@@ -378,7 +378,7 @@
 	sprintf(filename, "/dev/i2c/%d", i2cbus);
 	file = open(filename, O_RDWR);
 

Review line 384

Quoted 6 lines

(quoted diff) +1 -1

-	if (file < 0 && errno == ENOENT) {
+	if (file < 0 && (errno == ENOENT || errno == ENOTDIR)) {
 		sprintf(filename, "/dev/i2c-%d", i2cbus);
 		file = open(filename, O_RDWR);
 	}

I've never seen the case where /dev/i2c exists but isn't a directory (which is why that case isn't handled.) How did you get in this situation? It doesn't make much sense IMHO.

That being said, we can certainly apply your patch if it makes you happy, I don't see it as being terribly useful, but it also shouldn't hurt.