[PATCH libdrm 1/2] tests/tegra: add openclose test to check-target

[PATCH libdrm 1/2] tests/tegra: add openclose test to check-target

[Erik Faye-Lund]

This makes the test run under the 'make check'-taget.

Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
---
 tests/tegra/Makefile.am | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/tests/tegra/Makefile.am b/tests/tegra/Makefile.am
index 8e625c8f..f8e2a146 100644
--- a/tests/tegra/Makefile.am
+++ b/tests/tegra/Makefile.am
@@ -9,5 +9,7 @@ LDADD = \
 	../../tegra/libdrm_tegra.la \
 	../../libdrm.la
 
-noinst_PROGRAMS = \
+TESTS = \
 	openclose
+
+check_PROGRAMS = $(TESTS)
-- 
2.12.1

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH libdrm 2/2] tegra: update symbol-check

[Erik Faye-Lund]

I get a few more symbols in my build tegra-libraries, so let's
include these in the whitelist as well.

While we're at it, update the comment at the top.

Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
---
 tegra/tegra-symbol-check | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/tegra/tegra-symbol-check b/tegra/tegra-symbol-check
index 40208311..420469f4 100755
--- a/tegra/tegra-symbol-check
+++ b/tegra/tegra-symbol-check
@@ -1,11 +1,14 @@
 #!/bin/bash
 
-# The following symbols (past the first five) are taken from the public headers.
-# A list of the latter should be available Makefile.sources/LIBDRM_FREEDRENO_H_FILES
+# The following symbols (past the first nine) are taken from tegra.h.
 
 FUNCS=$(nm -D --format=bsd --defined-only ${1-.libs/libdrm_tegra.so} | awk '{print $3}'| while read func; do
 ( grep -q "^$func$" || echo $func )  <<EOF
+__bss_end__
+__bss_start__
 __bss_start
+__end__
+_bss_end__
 _edata
 _end
 _fini
-- 
2.12.1

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH libdrm 1/2] tests/tegra: add openclose test to check-target

[Emil Velikov]

Hi Erik,

On 29 March 2017 at 23:24, Erik Faye-Lund <kusmabite@gmail.com> wrote:
> This makes the test run under the 'make check'-taget.
>
The test doesn't do anything useful but open/close the tegra device.
Do you see much value in it or we can drop it all together ?
Thierry how do you feel on the topic ?

-Emil
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH libdrm 2/2] tegra: update symbol-check

[Emil Velikov]

On 29 March 2017 at 23:24, Erik Faye-Lund <kusmabite@gmail.com> wrote:
> I get a few more symbols in my build tegra-libraries, so let's
> include these in the whitelist as well.
>
> While we're at it, update the comment at the top.
>
> Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
R-B and pushed to master.

Out of curiosity:
What platform are you using that introduces these - musl, newer glibc, other ?

Would be great if we can strip platform specific symbols from the list
- do you have any ideas how we can do that ?
We could omit any symbols that start with __ but that does not sound too robust.

Thanks
Emil
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH libdrm 2/2] tegra: update symbol-check

[Erik Faye-Lund]

On Mon, Apr 3, 2017 at 6:59 PM, Emil Velikov <emil.l.velikov@gmail.com> wrote:
> On 29 March 2017 at 23:24, Erik Faye-Lund <kusmabite@gmail.com> wrote:
>> I get a few more symbols in my build tegra-libraries, so let's
>> include these in the whitelist as well.
>>
>> While we're at it, update the comment at the top.
>>
>> Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
> R-B and pushed to master.
>
> Out of curiosity:
> What platform are you using that introduces these - musl, newer glibc, other ?

I'm using glibc 2.25 (in particular, 2.25-1 from Arch Linux).

> Would be great if we can strip platform specific symbols from the list
> - do you have any ideas how we can do that ?
> We could omit any symbols that start with __ but that does not sound too robust.

I think ignoring anything that starts with both single and double
underscore would make sense, as those symbols are reserved by the
compiler and "implementation" (typically libc etc). Our API shouldn't
export these symbols in the first place, and I suspect code-review is
sufficient to ensure this.

But either way, I'd prefer to just get the test working again first.
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH libdrm 2/2] tegra: update symbol-check

[Emil Velikov]

On 3 April 2017 at 21:55, Erik Faye-Lund <kusmabite@gmail.com> wrote:
> On Mon, Apr 3, 2017 at 6:59 PM, Emil Velikov <emil.l.velikov@gmail.com> wrote:
>> On 29 March 2017 at 23:24, Erik Faye-Lund <kusmabite@gmail.com> wrote:
>>> I get a few more symbols in my build tegra-libraries, so let's
>>> include these in the whitelist as well.
>>>
>>> While we're at it, update the comment at the top.
>>>
>>> Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
>> R-B and pushed to master.
>>
>> Out of curiosity:
>> What platform are you using that introduces these - musl, newer glibc, other ?
>
> I'm using glibc 2.25 (in particular, 2.25-1 from Arch Linux).
>
Using up-to date Arch with glibc 2.25-1 and symbols are not there.
Perhaps you're using the ARM one ... either way it doesn't matter that
much.

>> Would be great if we can strip platform specific symbols from the list
>> - do you have any ideas how we can do that ?
>> We could omit any symbols that start with __ but that does not sound too robust.
>
> I think ignoring anything that starts with both single and double
> underscore would make sense, as those symbols are reserved by the
> compiler and "implementation" (typically libc etc). Our API shouldn't
> export these symbols in the first place, and I suspect code-review is
> sufficient to ensure this.
>
> But either way, I'd prefer to just get the test working again first.

Agreed. The patch is in master and part of 2.4.77

-Emil
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel