Re: [libvirt] [PATCH V2] libxl: don't overwrite error from virNetSocketNewConnectTCP()

Re: [libvirt] [PATCH V2] libxl: don't overwrite error from virNetSocketNewConnectTCP()

[John Ferlan]

On 09/03/2015 01:40 PM, Jim Fehlig wrote:
> Remove redundant error reporting in libxlDomainMigrationPerform().
> virNetSocketNewConnectTCP() is perfectly capable of reporting
> sensible errors.
> 
> Signed-off-by: Jim Fehlig <jfehlig@suse.com>
> ---
> 
> V2:
> Actually try to compile the code and find saved_errno is no
> longer used - remove it.
> 
>  src/libxl/libxl_migration.c | 7 +------
>  1 file changed, 1 insertion(+), 6 deletions(-)
> 

ACK

John

Also noteworthy that virNetSocketNewConnectTCP could overwrite errno
when freeaddrinfo is run in error: or when VIR_FORCE_CLOSE() did it's
thing...

Re: [libvirt] [PATCH V2] libxl: don't overwrite error from virNetSocketNewConnectTCP()

[Jim Fehlig]

On 09/03/2015 02:55 PM, John Ferlan wrote:
>
> On 09/03/2015 01:40 PM, Jim Fehlig wrote:
>> Remove redundant error reporting in libxlDomainMigrationPerform().
>> virNetSocketNewConnectTCP() is perfectly capable of reporting
>> sensible errors.
>>
>> Signed-off-by: Jim Fehlig <jfehlig@suse.com>
>> ---
>>
>> V2:
>> Actually try to compile the code and find saved_errno is no
>> longer used - remove it.
>>
>>   src/libxl/libxl_migration.c | 7 +------
>>   1 file changed, 1 insertion(+), 6 deletions(-)
>>
> ACK

Thanks! Pushed now.

Regards,
Jim

Re: [libvirt] [PATCH V2] libxl: don't overwrite error from virNetSocketNewConnectTCP()

[Michal Privoznik]

On 03.09.2015 22:55, John Ferlan wrote:
> 
> 
> On 09/03/2015 01:40 PM, Jim Fehlig wrote:
>> Remove redundant error reporting in libxlDomainMigrationPerform().
>> virNetSocketNewConnectTCP() is perfectly capable of reporting
>> sensible errors.
>>
>> Signed-off-by: Jim Fehlig <jfehlig@suse.com>
>> ---
>>
>> V2:
>> Actually try to compile the code and find saved_errno is no
>> longer used - remove it.
>>
>>  src/libxl/libxl_migration.c | 7 +------
>>  1 file changed, 1 insertion(+), 6 deletions(-)
>>
> 
> ACK
> 
> John
> 
> Also noteworthy that virNetSocketNewConnectTCP could overwrite errno
> when freeaddrinfo is run in error: or when VIR_FORCE_CLOSE() did it's
> thing...
> 

I don't think that either of those overwrites errno. Here's
freeaddrinfo() impl that I've found in glibc:

void
freeaddrinfo (struct addrinfo *ai)
{
  struct addrinfo *p;

  while (ai != NULL)
    {
      p = ai;
      ai = ai->ai_next;
      free (p->ai_canonname);
      free (p);
    }
}

And VIR_FORCE_CLOSE() preserves errno too. But nevertheless, ACK stands.

Michal