[dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[Pablo de Lara]

DPDK does not build on gcc 4.4, as it complains due to
strict-aliasing rules in virtual_pmd and link_bonding_mode4 tests,
with no errors in next gcc versions:

CC virtual_pmd.o
CC test_link_bonding_mode4.o
cc1: warnings being treated as errors
/root/dpdk/app/test/test_link_bonding_mode4.c: In function ‘test_mode4_expired’:
/root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error: dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
/root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error: dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
...
cc1: warnings being treated as errors
/root/dpdk/app/test/virtual_pmd.c: In function ‘virtual_ethdev_stats_reset’:
/root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error: dereferencing pointer ‘pkt.338’ does break strict-aliasing rules
/root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error: dereferencing pointer ‘pkt.338’ does break strict-aliasing rules

This patch fixes the compilation by disabling
these warnings.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
---
 app/test/Makefile |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/app/test/Makefile b/app/test/Makefile
index 4aca77c..0d62d07 100644
--- a/app/test/Makefile
+++ b/app/test/Makefile
@@ -158,6 +158,11 @@ CFLAGS_test_memcpy_perf.o += -fno-var-tracking-assignments
 endif
 endif
 
+# Disable warnings of no-strict-aliasing in virtual_pmd
+# and test_link_bonding_mode4 tests
+CFLAGS_virtual_pmd.o += -Wno-strict-aliasing
+CFLAGS_test_link_bonding_mode4.o += -Wno-strict-aliasing
+
 # this application needs libraries first
 DEPDIRS-y += lib
 
-- 
1.7.4.1

Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[Qiu, Michael]

On 3/27/2015 3:18 AM, Pablo de Lara wrote:
> DPDK does not build on gcc 4.4, as it complains due to

Why? Will gcc 4.4 not supported?

Thanks,
Michael
> strict-aliasing rules in virtual_pmd and link_bonding_mode4 tests,
> with no errors in next gcc versions:
>
> CC virtual_pmd.o
> CC test_link_bonding_mode4.o
> cc1: warnings being treated as errors
> /root/dpdk/app/test/test_link_bonding_mode4.c: In function ‘test_mode4_expired’:
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error: dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error: dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
> ...
> cc1: warnings being treated as errors
> /root/dpdk/app/test/virtual_pmd.c: In function ‘virtual_ethdev_stats_reset’:
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error: dereferencing pointer ‘pkt.338’ does break strict-aliasing rules
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error: dereferencing pointer ‘pkt.338’ does break strict-aliasing rules
>
> This patch fixes the compilation by disabling
> these warnings.
>
> Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
> ---
>  app/test/Makefile |    5 +++++
>  1 files changed, 5 insertions(+), 0 deletions(-)
>
> diff --git a/app/test/Makefile b/app/test/Makefile
> index 4aca77c..0d62d07 100644
> --- a/app/test/Makefile
> +++ b/app/test/Makefile
> @@ -158,6 +158,11 @@ CFLAGS_test_memcpy_perf.o += -fno-var-tracking-assignments
>  endif
>  endif
>  
> +# Disable warnings of no-strict-aliasing in virtual_pmd
> +# and test_link_bonding_mode4 tests
> +CFLAGS_virtual_pmd.o += -Wno-strict-aliasing
> +CFLAGS_test_link_bonding_mode4.o += -Wno-strict-aliasing
> +
>  # this application needs libraries first
>  DEPDIRS-y += lib
>  

Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[De Lara Guarch, Pablo]

> -----Original Message-----
> From: Qiu, Michael
> Sent: Friday, March 27, 2015 8:41 AM
> To: De Lara Guarch, Pablo; dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings
> 
> On 3/27/2015 3:18 AM, Pablo de Lara wrote:
> > DPDK does not build on gcc 4.4, as it complains due to
> 
> Why? Will gcc 4.4 not supported?

Afaik, we still support gcc 4.4, since it is part of RHEL 6.5.

Thanks,
Pablo
> 
> Thanks,
> Michael
> > strict-aliasing rules in virtual_pmd and link_bonding_mode4 tests,
> > with no errors in next gcc versions:
> >
> > CC virtual_pmd.o
> > CC test_link_bonding_mode4.o
> > cc1: warnings being treated as errors
> > /root/dpdk/app/test/test_link_bonding_mode4.c: In function
> 'test_mode4_expired':
> > /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error:
> dereferencing pointer 'pkt.409' does break strict-aliasing rules
> > /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error:
> dereferencing pointer 'pkt.409' does break strict-aliasing rules
> > ...
> > cc1: warnings being treated as errors
> > /root/dpdk/app/test/virtual_pmd.c: In function
> 'virtual_ethdev_stats_reset':
> > /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error:
> dereferencing pointer 'pkt.338' does break strict-aliasing rules
> > /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error:
> dereferencing pointer 'pkt.338' does break strict-aliasing rules
> >
> > This patch fixes the compilation by disabling
> > these warnings.
> >
> > Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
> > ---
> >  app/test/Makefile |    5 +++++
> >  1 files changed, 5 insertions(+), 0 deletions(-)
> >
> > diff --git a/app/test/Makefile b/app/test/Makefile
> > index 4aca77c..0d62d07 100644
> > --- a/app/test/Makefile
> > +++ b/app/test/Makefile
> > @@ -158,6 +158,11 @@ CFLAGS_test_memcpy_perf.o += -fno-var-
> tracking-assignments
> >  endif
> >  endif
> >
> > +# Disable warnings of no-strict-aliasing in virtual_pmd
> > +# and test_link_bonding_mode4 tests
> > +CFLAGS_virtual_pmd.o += -Wno-strict-aliasing
> > +CFLAGS_test_link_bonding_mode4.o += -Wno-strict-aliasing
> > +
> >  # this application needs libraries first
> >  DEPDIRS-y += lib
> >

Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[Thomas Monjalon]

2015-03-26 19:17, Pablo de Lara:
> DPDK does not build on gcc 4.4, as it complains due to
> strict-aliasing rules in virtual_pmd and link_bonding_mode4 tests,
> with no errors in next gcc versions:

Do you mean it's not possible to fix it in the code?
You are disabling some checks which may be relevant.

> CC virtual_pmd.o
> CC test_link_bonding_mode4.o
> cc1: warnings being treated as errors
> /root/dpdk/app/test/test_link_bonding_mode4.c: In function ‘test_mode4_expired’:
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error: dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error: dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
> ...
> cc1: warnings being treated as errors
> /root/dpdk/app/test/virtual_pmd.c: In function ‘virtual_ethdev_stats_reset’:
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error: dereferencing pointer ‘pkt.338’ does break strict-aliasing rules
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error: dereferencing pointer ‘pkt.338’ does break strict-aliasing rules
> 
> This patch fixes the compilation by disabling
> these warnings.
> 
> Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
> ---
>  app/test/Makefile |    5 +++++
>  1 files changed, 5 insertions(+), 0 deletions(-)
> 
> diff --git a/app/test/Makefile b/app/test/Makefile
> index 4aca77c..0d62d07 100644
> --- a/app/test/Makefile
> +++ b/app/test/Makefile
> @@ -158,6 +158,11 @@ CFLAGS_test_memcpy_perf.o += -fno-var-tracking-assignments
>  endif
>  endif
>  
> +# Disable warnings of no-strict-aliasing in virtual_pmd
> +# and test_link_bonding_mode4 tests
> +CFLAGS_virtual_pmd.o += -Wno-strict-aliasing
> +CFLAGS_test_link_bonding_mode4.o += -Wno-strict-aliasing
> +
>  # this application needs libraries first
>  DEPDIRS-y += lib
>  
> 

Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[De Lara Guarch, Pablo]

> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Friday, March 27, 2015 9:20 AM
> To: De Lara Guarch, Pablo
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings
> 
> 2015-03-26 19:17, Pablo de Lara:
> > DPDK does not build on gcc 4.4, as it complains due to
> > strict-aliasing rules in virtual_pmd and link_bonding_mode4 tests,
> > with no errors in next gcc versions:
> 
> Do you mean it's not possible to fix it in the code?
> You are disabling some checks which may be relevant.

I mean it is possible to fix, but probably not worth it, considering that
newer gcc versions have no issues (and gcc 4.4 is from 2011).
I have seen that there are other libraries that disable these warnings as well,
so I assumed we could do the same here.

Pablo
> 
> > CC virtual_pmd.o
> > CC test_link_bonding_mode4.o
> > cc1: warnings being treated as errors
> > /root/dpdk/app/test/test_link_bonding_mode4.c: In function
> ‘test_mode4_expired’:
> > /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error:
> dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
> > /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error:
> dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
> > ...
> > cc1: warnings being treated as errors
> > /root/dpdk/app/test/virtual_pmd.c: In function
> ‘virtual_ethdev_stats_reset’:
> > /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error:
> dereferencing pointer ‘pkt.338’ does break strict-aliasing rules
> > /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error:
> dereferencing pointer ‘pkt.338’ does break strict-aliasing rules
> >
> > This patch fixes the compilation by disabling
> > these warnings.
> >
> > Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
> > ---
> >  app/test/Makefile |    5 +++++
> >  1 files changed, 5 insertions(+), 0 deletions(-)
> >
> > diff --git a/app/test/Makefile b/app/test/Makefile
> > index 4aca77c..0d62d07 100644
> > --- a/app/test/Makefile
> > +++ b/app/test/Makefile
> > @@ -158,6 +158,11 @@ CFLAGS_test_memcpy_perf.o += -fno-var-
> tracking-assignments
> >  endif
> >  endif
> >
> > +# Disable warnings of no-strict-aliasing in virtual_pmd
> > +# and test_link_bonding_mode4 tests
> > +CFLAGS_virtual_pmd.o += -Wno-strict-aliasing
> > +CFLAGS_test_link_bonding_mode4.o += -Wno-strict-aliasing
> > +
> >  # this application needs libraries first
> >  DEPDIRS-y += lib
> >
> >
> 

Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[Thomas Monjalon]

2015-03-27 09:29, De Lara Guarch, Pablo:
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > 2015-03-26 19:17, Pablo de Lara:
> > > DPDK does not build on gcc 4.4, as it complains due to
> > > strict-aliasing rules in virtual_pmd and link_bonding_mode4 tests,
> > > with no errors in next gcc versions:
> > 
> > Do you mean it's not possible to fix it in the code?
> > You are disabling some checks which may be relevant.
> 
> I mean it is possible to fix, but probably not worth it, considering that
> newer gcc versions have no issues (and gcc 4.4 is from 2011).
> I have seen that there are other libraries that disable these warnings as well,
> so I assumed we could do the same here.

It is preferred to avoid doing it.

Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[De Lara Guarch, Pablo]

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
> Sent: Friday, March 27, 2015 10:01 AM
> To: De Lara Guarch, Pablo
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings
> 
> 2015-03-27 09:29, De Lara Guarch, Pablo:
> > From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > > 2015-03-26 19:17, Pablo de Lara:
> > > > DPDK does not build on gcc 4.4, as it complains due to
> > > > strict-aliasing rules in virtual_pmd and link_bonding_mode4 tests,
> > > > with no errors in next gcc versions:
> > >
> > > Do you mean it's not possible to fix it in the code?
> > > You are disabling some checks which may be relevant.
> >
> > I mean it is possible to fix, but probably not worth it, considering that
> > newer gcc versions have no issues (and gcc 4.4 is from 2011).
> > I have seen that there are other libraries that disable these warnings as
> well,
> > so I assumed we could do the same here.
> 
> It is preferred to avoid doing it.

Actually, I just realized that I missed checking the gcc version, so I was disabling the warnings in all versions.
I can send a v2 with that check, so we will still take into account those warnings for newer versions,
and just ignore it in gcc 4.4.

Thanks,
Pablo

[dpdk-dev] [PATCH v2 0/2] test: fix strict aliasing rule errors in gcc 4.4

[Pawel Wodkowski]

Fix strict aliasing rule error seen in gcc 4.4 in virtual pmd and mode 4 link
bonding mode 4 test.

changes v2:
 - fix error in code instead disabling compiler warning.

Pawel Wodkowski (2):
  test: fix strict aliasing rule error in virtual pmd
  test: fix strict aliasing rule error in link bonding mode 4 test

 app/test/test_link_bonding_mode4.c |  5 +++--
 app/test/virtual_pmd.c             | 10 +++++-----
 2 files changed, 8 insertions(+), 7 deletions(-)

-- 
1.9.1

[dpdk-dev] [PATCH v2 1/2] test: fix strict aliasing rule error in virtual pmd

[Pawel Wodkowski]

Fix strict aliasing rule error seen in gcc 4.4

Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
---
 app/test/virtual_pmd.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/app/test/virtual_pmd.c b/app/test/virtual_pmd.c
index 9b07ab1..f163562 100644
--- a/app/test/virtual_pmd.c
+++ b/app/test/virtual_pmd.c
@@ -75,15 +75,15 @@ virtual_ethdev_start_fail(struct rte_eth_dev *eth_dev __rte_unused)
 }
 static void  virtual_ethdev_stop(struct rte_eth_dev *eth_dev __rte_unused)
 {
-	struct rte_mbuf *pkt = NULL;
+	void *pkt = NULL;
 	struct virtual_ethdev_private *prv = eth_dev->data->dev_private;
 
 	eth_dev->data->dev_link.link_status = 0;
 	eth_dev->data->dev_started = 0;
-	while (rte_ring_dequeue(prv->rx_queue, (void **)&pkt) != -ENOENT)
+	while (rte_ring_dequeue(prv->rx_queue, &pkt) != -ENOENT)
 		rte_pktmbuf_free(pkt);
 
-	while (rte_ring_dequeue(prv->tx_queue, (void **)&pkt) != -ENOENT)
+	while (rte_ring_dequeue(prv->tx_queue, &pkt) != -ENOENT)
 		rte_pktmbuf_free(pkt);
 }
 
@@ -223,9 +223,9 @@ static void
 virtual_ethdev_stats_reset(struct rte_eth_dev *dev)
 {
 	struct virtual_ethdev_private *dev_private = dev->data->dev_private;
-	struct rte_mbuf *pkt = NULL;
+	void *pkt = NULL;
 
-	while (rte_ring_dequeue(dev_private->tx_queue, (void **)&pkt) == -ENOBUFS)
+	while (rte_ring_dequeue(dev_private->tx_queue, &pkt) == -ENOBUFS)
 			rte_pktmbuf_free(pkt);
 
 	/* Reset internal statistics */
-- 
1.9.1

[dpdk-dev] [PATCH v2 2/2] test: fix strict aliasing rule error in link bonding mode 4 test

[Pawel Wodkowski]

Fix strict aliasing rule error seen in gcc 4.4

Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
---
 app/test/test_link_bonding_mode4.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/app/test/test_link_bonding_mode4.c b/app/test/test_link_bonding_mode4.c
index c35129f..02380f9 100644
--- a/app/test/test_link_bonding_mode4.c
+++ b/app/test/test_link_bonding_mode4.c
@@ -1181,7 +1181,6 @@ test_mode4_expired(void)
 {
 	struct slave_conf *slave, *exp_slave = NULL;
 	struct rte_mbuf *pkts[MAX_PKT_BURST];
-	struct rte_mbuf *pkt = NULL;
 	int retval;
 	uint32_t old_delay;
 
@@ -1239,7 +1238,9 @@ test_mode4_expired(void)
 			/* Remove replay for slave that supose to be expired. */
 			if (slave == exp_slave) {
 				while (rte_ring_count(slave->rx_queue) > 0) {
-					rte_ring_dequeue(slave->rx_queue, (void **)&pkt);
+					void *pkt = NULL;
+
+					rte_ring_dequeue(slave->rx_queue, &pkt);
 					rte_pktmbuf_free(pkt);
 				}
 			}
-- 
1.9.1

Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[Stephen Hemminger]

On Thu, 26 Mar 2015 19:17:35 +0000
Pablo de Lara <pablo.de.lara.guarch@intel.com> wrote:

> +# Disable warnings of no-strict-aliasing in virtual_pmd
> +# and test_link_bonding_mode4 tests
> +CFLAGS_virtual_pmd.o += -Wno-strict-aliasing
> +CFLAGS_test_link_bonding_mode4.o += -Wno-strict-aliasing
> +
>  

There is a way to do it only for a specific (broken)
Gcc version using if's

Re: [dpdk-dev] [PATCH v2 1/2] test: fix strict aliasing rule error in virtual pmd

[Stephen Hemminger]

On Fri, 27 Mar 2015 11:56:00 +0100
Pawel Wodkowski <pawelx.wodkowski@intel.com> wrote:

> Fix strict aliasing rule error seen in gcc 4.4
> 
> Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
> ---
>  app/test/virtual_pmd.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/app/test/virtual_pmd.c b/app/test/virtual_pmd.c
> index 9b07ab1..f163562 100644
> --- a/app/test/virtual_pmd.c
> +++ b/app/test/virtual_pmd.c
> @@ -75,15 +75,15 @@ virtual_ethdev_start_fail(struct rte_eth_dev *eth_dev __rte_unused)
>  }
>  static void  virtual_ethdev_stop(struct rte_eth_dev *eth_dev __rte_unused)
>  {
> -	struct rte_mbuf *pkt = NULL;
> +	void *pkt = NULL;
>  	struct virtual_ethdev_private *prv = eth_dev->data->dev_private;
>  
>  	eth_dev->data->dev_link.link_status = 0;
>  	eth_dev->data->dev_started = 0;
> -	while (rte_ring_dequeue(prv->rx_queue, (void **)&pkt) != -ENOENT)
> +	while (rte_ring_dequeue(prv->rx_queue, &pkt) != -ENOENT)
>  		rte_pktmbuf_free(pkt);
>  
> -	while (rte_ring_dequeue(prv->tx_queue, (void **)&pkt) != -ENOENT)
> +	while (rte_ring_dequeue(prv->tx_queue, &pkt) != -ENOENT)
>  		rte_pktmbuf_free(pkt);
>  }

Losing type information doesn't make this better!

Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[De Lara Guarch, Pablo]

> -----Original Message-----
> From: Stephen Hemminger [mailto:stephen@networkplumber.org]
> Sent: Friday, March 27, 2015 3:46 PM
> To: De Lara Guarch, Pablo
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings
> 
> On Thu, 26 Mar 2015 19:17:35 +0000
> Pablo de Lara <pablo.de.lara.guarch@intel.com> wrote:
> 
> > +# Disable warnings of no-strict-aliasing in virtual_pmd
> > +# and test_link_bonding_mode4 tests
> > +CFLAGS_virtual_pmd.o += -Wno-strict-aliasing
> > +CFLAGS_test_link_bonding_mode4.o += -Wno-strict-aliasing
> > +
> >
> 
> There is a way to do it only for a specific (broken)
> Gcc version using if's

I know, I sent another email, saying that I missed an if to check the gcc version.
Anyway, Pawel has sent another patch to fix the issue.
Not sure which fix the community prefers.

Thanks,
Pablo

[dpdk-dev] [PATCH v2] DPDK does not build on gcc 4.4, as it complains due to strict-aliasing rules in virtual_pmd and link_bonding_mode4 tests, with no errors in next gcc versions:

[Pablo de Lara]

CC virtual_pmd.o
CC test_link_bonding_mode4.o
cc1: warnings being treated as errors
/root/dpdk/app/test/test_link_bonding_mode4.c: In function ‘test_mode4_expired’:
/root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error: dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
/root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error: dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
...
cc1: warnings being treated as errors
/root/dpdk/app/test/virtual_pmd.c: In function ‘virtual_ethdev_stats_reset’:
/root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error: dereferencing pointer ‘pkt.338’ does break strict-aliasing rules
/root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error: dereferencing pointer ‘pkt.338’ does break strict-aliasing rules

This patch fixes the compilation by disabling
these warnings.

Changes in v2:

Added missing gcc version check (checking if gcc 4.4 is present)

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
---
 app/test/Makefile |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/app/test/Makefile b/app/test/Makefile
index 4aca77c..9639a71 100644
--- a/app/test/Makefile
+++ b/app/test/Makefile
@@ -158,6 +158,15 @@ CFLAGS_test_memcpy_perf.o += -fno-var-tracking-assignments
 endif
 endif
 
+# Disable warnings of no-strict-aliasing in virtual_pmd
+# and test_link_bonding_mode4 tests
+ifeq ($(CC), gcc)
+ifeq ($(shell test $(GCC_VERSION) -eq 44 && echo 1), 1)
+CFLAGS_virtual_pmd.o += -Wno-strict-aliasing
+CFLAGS_test_link_bonding_mode4.o += -Wno-strict-aliasing
+endif
+endif
+
 # this application needs libraries first
 DEPDIRS-y += lib
 
-- 
1.7.4.1

Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings

[De Lara Guarch, Pablo]

> -----Original Message-----
> From: De Lara Guarch, Pablo
> Sent: Friday, March 27, 2015 10:11 AM
> To: Thomas Monjalon
> Cc: dev@dpdk.org
> Subject: RE: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings
> 
> 
> 
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
> > Sent: Friday, March 27, 2015 10:01 AM
> > To: De Lara Guarch, Pablo
> > Cc: dev@dpdk.org
> > Subject: Re: [dpdk-dev] [PATCH] test: Disable strict-aliasing warnings
> >
> > 2015-03-27 09:29, De Lara Guarch, Pablo:
> > > From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > > > 2015-03-26 19:17, Pablo de Lara:
> > > > > DPDK does not build on gcc 4.4, as it complains due to
> > > > > strict-aliasing rules in virtual_pmd and link_bonding_mode4 tests,
> > > > > with no errors in next gcc versions:
> > > >
> > > > Do you mean it's not possible to fix it in the code?
> > > > You are disabling some checks which may be relevant.
> > >
> > > I mean it is possible to fix, but probably not worth it, considering that
> > > newer gcc versions have no issues (and gcc 4.4 is from 2011).
> > > I have seen that there are other libraries that disable these warnings as
> > well,
> > > so I assumed we could do the same here.
> >
> > It is preferred to avoid doing it.
> 
> Actually, I just realized that I missed checking the gcc version, so I was
> disabling the warnings in all versions.
> I can send a v2 with that check, so we will still take into account those
> warnings for newer versions,
> and just ignore it in gcc 4.4.

I just sent a v2 of this patch. Pawel also sent another fix for this issue,
but Stephen had some concerns about it, so I just sent a v2 as an alternative,
so the community can decide between the two options.

> 
> Thanks,
> Pablo

Re: [dpdk-dev] [PATCH v2] DPDK does not build on gcc 4.4, as it complains due to strict-aliasing rules in virtual_pmd and link_bonding_mode4 tests, with no errors in next gcc versions:

[Thomas Monjalon]

2015-03-30 12:06, Pablo de Lara:
> CC virtual_pmd.o
> CC test_link_bonding_mode4.o
> cc1: warnings being treated as errors
> /root/dpdk/app/test/test_link_bonding_mode4.c: In function ‘test_mode4_expired’:
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error: dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error: dereferencing pointer ‘pkt.409’ does break strict-aliasing rules
> ...
> cc1: warnings being treated as errors
> /root/dpdk/app/test/virtual_pmd.c: In function ‘virtual_ethdev_stats_reset’:
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:645: error: dereferencing pointer ‘pkt.338’ does break strict-aliasing rules
> /root/dpdk/x86_64-native-linuxapp-gcc/include/rte_ring.h:729: error: dereferencing pointer ‘pkt.338’ does break strict-aliasing rules
> 
> This patch fixes the compilation by disabling
> these warnings.
> 
> Changes in v2:
> 
> Added missing gcc version check (checking if gcc 4.4 is present)
> 
> Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>

Pawel's proposal was chosen because it doesn't require handling special
compiler version.

Re: [dpdk-dev] [PATCH v2 0/2] test: fix strict aliasing rule errors in gcc 4.4

[Thomas Monjalon]

2015-03-27 11:55, Pawel Wodkowski:
> Fix strict aliasing rule error seen in gcc 4.4 in virtual pmd and mode 4 link
> bonding mode 4 test.
> 
> changes v2:
>  - fix error in code instead disabling compiler warning.
> 
> Pawel Wodkowski (2):
>   test: fix strict aliasing rule error in virtual pmd
>   test: fix strict aliasing rule error in link bonding mode 4 test
> 
>  app/test/test_link_bonding_mode4.c |  5 +++--
>  app/test/virtual_pmd.c             | 10 +++++-----
>  2 files changed, 8 insertions(+), 7 deletions(-)

Applied, thanks