Debian Bug report logs - #873608
uhd: NEON-related FTBFS on armhf

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: Adrian Bunk <bunk@debian.org>

To: Debian Bug Tracking System <submit@bugs.debian.org>

Subject: uhd FTBFS on armhf dues to NEON

Date: Tue, 29 Aug 2017 16:16:06 +0300

Source: uhd
Version: 3.10.2.0-1
Severity: serious

https://buildd.debian.org/status/fetch.php?pkg=uhd&arch=armhf&ver=3.10.2.0-1&stamp=1503874671&raw=0

...
In file included from /«PKGBUILDDIR»/host/lib/convert/convert_with_neon.cpp:20:0:
/usr/lib/gcc/arm-linux-gnueabihf/7/include/arm_neon.h: In member function 'virtual void __convert_fc32_1_sc16_item32_le_1_PRIORITY_SIMD::operator()(const input_type&, const output_type&, size_t)':
/usr/lib/gcc/arm-linux-gnueabihf/7/include/arm_neon.h:6740:1: error: inlining failed in call to always_inline 'float32x4_t vdupq_n_f32(float32_t)': target specific option mismatch
 vdupq_n_f32 (float32_t __a)
 ^~~~~~~~~~~
/«PKGBUILDDIR»/host/lib/convert/convert_with_neon.cpp:36:53: note: called from here
     float32x4_t Q0 = vdupq_n_f32(float(scale_factor));
                                                     ^
In file included from /«PKGBUILDDIR»/host/lib/convert/convert_with_neon.cpp:20:0:
/usr/lib/gcc/arm-linux-gnueabihf/7/include/arm_neon.h:10844:1: error: inlining failed in call to always_inline 'void vst1_s16(int16_t*, int16x4_t)': target specific option mismatch
 vst1_s16 (int16_t * __a, int16x4_t __b)
 ^~~~~~~~
/«PKGBUILDDIR»/host/lib/convert/convert_with_neon.cpp:65:17: note: called from here
         vst1_s16((reinterpret_cast<int16_t *>(&output[i+6])), D15);
         ~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from /«PKGBUILDDIR»/host/lib/convert/convert_with_neon.cpp:20:0:
/usr/lib/gcc/arm-linux-gnueabihf/7/include/arm_neon.h:9027:1: error: inlining failed in call to always_inline 'int16x4_t vrev32_s16(int16x4_t)': target specific option mismatch
 vrev32_s16 (int16x4_t __a)
 ^~~~~~~~~~
/«PKGBUILDDIR»/host/lib/convert/convert_with_neon.cpp:64:39: note: called from here
         int16x4_t D15 = vrev32_s16(D14);
                                       ^
In file included from /«PKGBUILDDIR»/host/lib/convert/convert_with_neon.cpp:20:0:
/usr/lib/gcc/arm-linux-gnueabihf/7/include/arm_neon.h:7552:1: error: inlining failed in call to always_inline 'int16x4_t vmovn_s32(int32x4_t)': target specific option mismatch
 vmovn_s32 (int32x4_t __a)
 ^~~~~~~~~
/«PKGBUILDDIR»/host/lib/convert/convert_with_neon.cpp:63:38: note: called from here
         int16x4_t D14 = vmovn_s32(Q13);
                                      ^
...



NEON is not part of the armhf port baseline.

Message #12 received at 873608@bugs.debian.org (full text, mbox, reply):

From: Paul Tagliamonte <paultag@debian.org>

To: 873608@bugs.debian.org, control@bugs.debian.org

Subject: patch

Date: Tue, 29 Aug 2017 16:10:30 -0400

[Message part 1 (text/plain, inline)]

tags 873608 + patch
thanks

Attached is a patch that will enable neon if arm_neon.h is present. I
didn't upstream this or anything, I figure you have a better
relationship with them, but this ought to fix the ftbfs.

I'm running a test build now, but it's past the old ftbfs point

  Paul

-- 

[neon.patch (text/x-diff, attachment)]

Message #19 received at 873608@bugs.debian.org (full text, mbox, reply):

From: Paul Tagliamonte <paultag@debian.org>

To: 873608@bugs.debian.org

Subject: Re: patch

Date: Tue, 29 Aug 2017 17:13:25 -0400

Whoops, I only just read the comment about this not being standard in
armhf. I misread that and also assumed it was standard (likely because,
as a Coretex-A* user, I always had it), so this patch will likely be
inappropriate (or incomplete). Maybe this patch plus an AND on an
ENABLE_NEON or something, which is disabled by default in the packaging.

  Paul


On Tue, Aug 29, 2017 at 04:10:30PM -0400, Paul Tagliamonte wrote:
> tags 873608 + patch
> thanks
> 
> Attached is a patch that will enable neon if arm_neon.h is present. I
> didn't upstream this or anything, I figure you have a better
> relationship with them, but this ought to fix the ftbfs.
> 
> I'm running a test build now, but it's past the old ftbfs point
> 
>   Paul
> 
> -- 



-- 

Message #24 received at 873608@bugs.debian.org (full text, mbox, reply):

From: "A. Maitland Bottoms" <bottoms@debian.org>

To: 873608@bugs.debian.org

Subject: uhd: NEON-related FTBFS on armhf

Date: Tue, 05 Sep 2017 18:23:23 -0500

So yeah, we cannot expect NEON on Debian armhf.

But I did not expect these compile time errors... I would hope that
one could at least compile some NEON code on Debian armhf, and then
maybe do some runtime conditional test to only execute NEON if it
is available.

This code has not compiled on Debian for a while, so I am bringing back
a debian-armhf-convert-without-neon patch I used before.

-Maitland

Message #29 received at 873608-close@bugs.debian.org (full text, mbox, reply):

From: bottoms@debian.org (A. Maitland Bottoms)

To: 873608-close@bugs.debian.org

Subject: Bug#873608: fixed in uhd 3.10.2.0-2

Date: Wed, 06 Sep 2017 00:20:16 +0000

Source: uhd
Source-Version: 3.10.2.0-2

We believe that the bug you reported is fixed in the latest version of
uhd, which is due to be installed in the Debian FTP archive.

A summary of the changes between this version and the previous one is
attached.

Thank you for reporting the bug, which will now be closed.  If you
have further comments please address them to 873608@bugs.debian.org,
and the maintainer will reopen the bug report if appropriate.

Debian distribution maintenance software
pp.
A. Maitland Bottoms <bottoms@debian.org> (supplier of updated uhd package)

(This message was generated automatically at their request; if you
believe that there is a problem with it please contact the archive
administrators by mailing ftpmaster@ftp-master.debian.org)


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Format: 1.8
Date: Tue, 05 Sep 2017 18:48:41 -0400
Source: uhd
Binary: uhd-host libuhd003.010.002 libuhd-dev
Architecture: source
Version: 3.10.2.0-2
Distribution: unstable
Urgency: medium
Maintainer: A. Maitland Bottoms <bottoms@debian.org>
Changed-By: A. Maitland Bottoms <bottoms@debian.org>
Description:
 libuhd-dev - universal hardware driver for Ettus Research products - headers
 libuhd003.010.002 - universal hardware driver for Ettus Research products - library
 uhd-host   - universal hardware driver for Ettus Research products - host apps
Closes: 873608
Changes:
 uhd (3.10.2.0-2) unstable; urgency=medium
 .
   * avoid gcc 7 and armhf neon for now (Closes: 873608)
     by bringing back debian-armhf-convert-without-neon patch
Checksums-Sha1:
 41106d8ad12e6676a1f221d372a98c65f9b5fe29 2721 uhd_3.10.2.0-2.dsc
 99034f931e1ef56822d5dea235aad9c489f97247 237544 uhd_3.10.2.0-2.debian.tar.xz
 d85b680bdc132c859e2e8f8c3291f82aa7fc3677 11939 uhd_3.10.2.0-2_amd64.buildinfo
Checksums-Sha256:
 81eaeb9b88692977515327ed9a52af6919398d5b76122e24ce8765d184fd7982 2721 uhd_3.10.2.0-2.dsc
 a327ecd38da2739f54b6319ff903229e58eefd2c477078810e4c16670a9ad78b 237544 uhd_3.10.2.0-2.debian.tar.xz
 f76e0cdd02753c8ba4fe74a0445b6841c5f92089cfae1a001e98119687770dc4 11939 uhd_3.10.2.0-2_amd64.buildinfo
Files:
 80a36a8c2d6e6c334d585b2d8098a231 2721 science optional uhd_3.10.2.0-2.dsc
 5de13ecf7f7534d30732f8e92f511b77 237544 science optional uhd_3.10.2.0-2.debian.tar.xz
 da4379f6a2bb2fbce1b9284ddd437407 11939 science optional uhd_3.10.2.0-2_amd64.buildinfo

-----BEGIN PGP SIGNATURE-----
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=9y1i
-----END PGP SIGNATURE-----

Send a report that this bug log contains spam.