[depends] Allow depends system to support armv7l #12474

hkjn commented at 1:12 PM on February 17, 2018: contributor

Credit for the actual patches goes to @theuni, I just tested them on an armv7l machine.

This change allows the depends system to build our dependencies on 32-bit ARM / armv7l architecture. I.e. with this patch, the following steps now builds fine:

$ uname -m
armv7l
$ ./autogen.sh
$ cd depends && make NO_QT=1 && cd ..
$ ./configure --prefix=$(pwd)/depends/armv7l-unknown-linux-gnueabihf
$ make

Without this patch, the cd depends && make NO_QT=1 command fails on armv7l.

fanquake added the label Build system on Feb 17, 2018

hkjn force-pushed on Feb 17, 2018

fanquake commented at 3:06 AM on February 19, 2018: member

Travis failure on the arm-linux-gnueabihf build:

Extracting openssl...
/home/travis/build/bitcoin/bitcoin/depends/sources/openssl-1.0.1k.tar.gz: OK
Preprocessing openssl...
Configuring openssl...
Configuring for 
Usage: Configure [no-<cipher> ...] [enable-<cipher> ...] [experimental-<cipher> ...] [-Dxxx] [-lxxx] [-Lxxx] [-fxxx] [-Kxxx] [no-hw-xxx|no-hw] [[no-]threads] [[no-]shared] [[no-]zlib|zlib-dynamic] [no-asm] [no-dso] [no-krb5] [sctp] [386] [--prefix=DIR] [--openssldir=OPENSSLDIR] [--with-xxx[=vvv]] [--test-sanity] os/compiler[:flags]
pick os/compiler from:
BC-32 BS2000-OSD BSD-generic32 BSD-generic64
<snip>
unixware-7-gcc vos-gcc vxworks-mips

NOTE: If in doubt, on Unix-ish systems use './config'.
make: *** [/home/travis/build/bitcoin/bitcoin/depends/work/build/arm-linux-gnueabihf/openssl/1.0.1k-e3fad431e69/./.stamp_configured] Error 1
make: Leaving directory `/home/travis/build/bitcoin/bitcoin/depends'

android-armv7 and linux-armv4 are listed in the snipped os/compiler options.

hkjn force-pushed on Feb 19, 2018

hkjn commented at 2:24 PM on February 19, 2018: contributor

Indeed @fanquake, sorry about that, I had messed up one line in depends/packages/openssl.mk. Repushed d79ee76, hopefully Travis will be happier this time..

theuni commented at 2:48 PM on February 19, 2018: member

Disclaimer: these were quick hacks to see what it would take to build on non-x86, I'd prefer a change that's a little more obvious. Mind testing this instead of the first chunk?

diff --git a/depends/Makefile b/depends/Makefile
index 0ddd348e53..14e94ba453 100644
--- a/depends/Makefile
+++ b/depends/Makefile
@@ -21,7 +21,6 @@ BUILD_ID_SALT ?= salt
 host:=$(BUILD)
 ifneq ($(HOST),)
 host:=$(HOST)
-host_toolchain:=$(HOST)-
 endif

 ifneq ($(DEBUG),)
diff --git a/depends/hosts/default.mk b/depends/hosts/default.mk
index 6f60d6b3fd..144e5f88b7 100644
--- a/depends/hosts/default.mk
+++ b/depends/hosts/default.mk
@@ -1,3 +1,7 @@
+ifneq ($(host),$(build))
+host_toolchain:=$(host)-
+endif
+
 default_host_CC = $(host_toolchain)gcc
 default_host_CXX = $(host_toolchain)g++
 default_host_AR = $(host_toolchain)ar

hkjn commented at 3:14 PM on February 19, 2018: contributor

Sure @theuni, see second commit f8719f8. (Let me know also if you'd like me to close the PR if you want to take the change instead; I'll admit that at this stage no changes to the build system are particularly obvious to me. :))

I'll squash commits into a single one (unless someone thinks there should be several) after review.

theuni commented at 7:45 PM on February 19, 2018: member

@hkjn thanks! This PR is fine. I assume the new change works fine for you? If so, looks good to me after squashing.

hkjn force-pushed on Feb 20, 2018

hkjn commented at 12:03 PM on February 20, 2018: contributor

@theuni yup, this patch still fixes build system when using depends locally for me.

I've squashed the commits.

[depends] Allow depends system to support armv7l ac91ea64c2

hkjn commented at 2:44 PM on February 20, 2018: contributor

The last Travis failure seems unrelated; it's for x86_64, and the failing test is mempool_limit.py.

I'll try rebasing to a later commit to trigger another build, as I'm not a member of the bitcoin org on github and can't start Travis builds directly..

hkjn force-pushed on Feb 20, 2018

theuni commented at 5:26 PM on February 20, 2018: member

Thanks! utACK ac91ea64c2b1daec8fee0bfde443c18cf87595ae @laanwj unfortunately this will cause PRs to rebuild depends. It may be helpful to merge changes like these on a Friday afternoon to help with Travis slow-downs.

laanwj added this to the milestone 0.16.1 on Feb 23, 2018

laanwj added the label Needs backport on Feb 23, 2018

laanwj commented at 4:12 PM on February 23, 2018: member

@theuni OK, as it is Friday afternoon (Dutch time) will merge it now. Also seems to make sense to backport to 0.16, and as 0.16.0 has been tagged I'll do that immediately too.

laanwj merged this on Feb 23, 2018

laanwj closed this on Feb 23, 2018

laanwj referenced this in commit 28b8061d9f on Feb 23, 2018

laanwj referenced this in commit b0f692f81e on Feb 23, 2018

laanwj removed the label Needs backport on Feb 23, 2018

hkjn deleted the branch on Feb 26, 2018

HashUnlimited referenced this in commit 0300f9b26c on Mar 16, 2018

UdjinM6 referenced this in commit 04d1671b9c on Jan 5, 2019

CryptoCentric referenced this in commit b0d26543f8 on Jul 11, 2019

DrahtBot locked this on Sep 8, 2021