Since the introduction of CMake 4 in several distributions, build errors
started occurring on those distributions for cmake packages that
included calls to cmake_minimum_required() or cmake_policy() with a
version older than 3.50 (see [1]).
To avoid backporting numerous individual fixes for affected packages,
commit [2] was previously applied to the LTS branch as a workaround.
Following further discussion (see [2][3]), that raised concerns about
policy changes between CMake versions that could lead to breaking builds
in non-obvious ways, a different approach was chosen.
This patch reverts commit [2] and instead of applying a global
compatibility variable, we now enforce building host-cmake when the host
system provides CMake 4 or newer.
[1] https://cmake.org/cmake/help/latest/release/4.0.html#deprecated-and-removed-features
[2] 70aac2d9e8 package/pkg-cmake.mk: force config version >=3.5
[3] https://lists.buildroot.org/pipermail/buildroot/2025-May/780262.html
[4] https://lists.buildroot.org/pipermail/buildroot/2025-June/780372.html
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Since the mbedtls bump to v3.6 [1] the libuhttpd fails to build with the
following error:
```
[ 8%] Building C object src/ssl/CMakeFiles/xssl.dir/mbedtls.c.o
.../buildroot/output/build/libuhttpd-3.14.1/src/ssl/mbedtls.c:52:10: fatal error: mbedtls/certs.h: No such file or directory
52 | #include <mbedtls/certs.h>
| ^~~~~~~~~~~~~~~~~
compilation terminated.
```
This error can be reproduced with the following config:
```
cat <<EOF >.config
BR2_arm=y
BR2_cortex_a7=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN=y
BR2_PACKAGE_MBEDTLS=y
BR2_PACKAGE_LIBUHTTPD=y
EOF
make olddefconfig
make
```
The compatibility with mbedtls v3 has been addressed upstream in the
zhaojh329/ssl project included as a submodule of libuhttpd [2].
This patch backport this upstream commit to be applied on the submodule
directory. This required adaptation of the line numbers (see [3]) and
renaming a function reference passed as parameter of
'mbedtls_pk_parse_keyfile' caused by the commit [4].
[1] 3481a9643f package/mbedtls: bump to version 3.6.3.1
[2] 28cc9b5d98
[3] 8092b5a490 (diff-fbc46fa2db83f8649ccf1f46c6a044473b7b228edc7d4c0f7cc04b5a879f6fb7)
[4] 0e7d2f73d7 (diff-fbc46fa2db83f8649ccf1f46c6a044473b7b228edc7d4c0f7cc04b5a879f6fb7R92)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 1a8e868623)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Since the mbedtls bump to v3.6 [1] the bmx7 package is failling on the
autobuilder:
```
/workdir/instance-0/output-1/host/bin/xtensa-buildroot-linux-uclibc-gcc -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -mlongcalls -mauto-litpools -Os -g3 -pedantic -W -Wall -Wstrict-prototypes -Wno-unused-parameter -Os -g3 -std=gnu99 -DGIT_REV=\"0\" -DAVL_5XLINKED -DDEBUG_MALLOC -DCORE_LIMIT=20000 -pedantic -W -Wall -Wstrict-prototypes -Wno-unused-parameter -Os -g3 -std=gnu99 -DGIT_REV=\"0\" -DAVL_5XLINKED -DDEBUG_MALLOC -DCORE_LIMIT=20000 -c crypt.c -o crypt.o
crypt.c:66:10: fatal error: mbedtls/compat-1.3.h: No such file or directory
66 | #include "mbedtls/compat-1.3.h"
| ^~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
```
This patch includes a set of upstream patches that address the
compatibility with the v3.6 of mbedtls.
[1] 3481a9643f package/mbedtls: bump to version 3.6.3.1
Fixes: https://autobuild.buildroot.org/results/b77/b776e34d1c5bc3904ea7138bd6c4ac17a1f0fd34/
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit c32230fe35)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Since the mbedtls bump to v3.6 [1] the shadowsocks-libev package is
failling on the autobuilder:
```
checking for mbedtls_cipher_setup in -lmbedcrypto... yes
checking whether mbedtls supports Cipher Feedback mode or not... configure: error: MBEDTLS_CIPHER_MODE_CFB required
make: *** [package/pkg-generic.mk:263: /workdir/instance-0/output-1/build/shadowsocks-libev-3.3.5/.stamp_configured] Error 1
make: Leaving directory '/workdir/instance-0/buildroot'
```
This is due to the breaking changes in the mbedtls API with the version
bump.
This patch adds the upstream patch [2] that address this issue by verifying
conditionally the version of mbedtls we are running on to make the API
calls and includes.
[1] 3481a9643f package/mbedtls: bump to version 3.6.3.1
[2] 9afa3cacf9#
Fixes: https://autobuild.buildroot.org/results/070/070581d95f2739cee3b4cb8252639dd92b5a8421
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 78198bc0f3)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
When SIGINT is issued for a package test run, it will abort the active
toolchain run then proceed to the next. If a user is running the entire
default toolchain set (`-a`), they can be required to invoke SIGINT
multiple times to stop a run.
This commit uses a SIGINT hook to flag a shutdown state and stop further
attempts to run anymore toolchain tests.
Signed-off-by: James Knight <git@jdknight.me>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 8f09106e81)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
When running check-package before completing commits for a change, if
any files are setup for removal, check-package will throw
FileNotFoundError exceptions instead of generating a warning state. For
example:
$ utils/docker-run make check-package
Traceback (most recent call last):
...
FileNotFoundError: [Errno 2] No such file or directory: 'package/.../0001-some-removed-patch.patch'
make: *** [Makefile:1264: check-package] Error 1
This commit will now catch FileNotFoundError and populate a warning
message:
$ utils/docker-run make check-package
package/.../0001-some-removed-patch.patch: missing; unstaged file removal?
package/.../0002-another-removed-patch.patch: missing; unstaged file removal?
427843 lines processed
3 warnings generated
make: *** [Makefile:1264: check-package] Error 1
Signed-off-by: James Knight <git@jdknight.me>
Reviewed-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit c41a06bbd9)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Import all security patches from OpenEmbedded for libsoup.
This fixes the following 18 known vulnerabilities:
- CVE-2024-52530:
GNOME libsoup before 3.6.0 allows HTTP request smuggling in some
configurations because '\0' characters at the end of header names are
ignored, i.e., a "Transfer-Encoding\0: chunked" header is treated the
same as a "Transfer-Encoding: chunked" header.
https://www.cve.org/CVERecord?id=CVE-2024-52530
- CVE-2024-52531:
GNOME libsoup before 3.6.1 allows a buffer overflow in applications that
perform conversion to UTF-8 in soup_header_parse_param_list_strict.
There is a plausible way to reach this remotely via
soup_message_headers_get_content_type (e.g., an application may want to
retrieve the content type of a request or response).
https://www.cve.org/CVERecord?id=CVE-2024-52531
- CVE-2024-52532:
GNOME libsoup before 3.6.1 has an infinite loop, and memory consumption.
During the reading of certain patterns of WebSocket data from clients.
https://www.cve.org/CVERecord?id=CVE-2024-52532
- CVE-2025-2784:
Libsoup: heap buffer over-read in `skip_insignificant_space`
when sniffing content
https://www.cve.org/CVERecord?id=CVE-2025-2784
- CVE-2025-4476:
Libsoup: null pointer dereference in libsoup may lead to denial of service
https://www.cve.org/CVERecord?id=CVE-2025-4476
- CVE-2025-4948:
Libsoup: integer underflow in soup_multipart_new_from_message() leading to
denial of service in libsoup
https://www.cve.org/CVERecord?id=CVE-2025-4948
- CVE-2025-4969:
Libsoup: off-by-one out-of-bounds read in find_boundary() in soup-multipart.c
https://www.cve.org/CVERecord?id=CVE-2025-4969
- CVE-2025-32050:
Libsoup: integer overflow in append_param_quoted
https://www.cve.org/CVERecord?id=CVE-2025-32050
- CVE-2025-32052:
Libsoup: heap buffer overflow in sniff_unknown()
https://www.cve.org/CVERecord?id=CVE-2025-32052
- CVE-2025-32053:
Libsoup: heap buffer overflows in sniff_feed_or_html() and
skip_insignificant_space()
https://www.cve.org/CVERecord?id=CVE-2025-32053
- CVE-2025-32906:
Libsoup: out of bounds reads in soup_headers_parse_request()
https://www.cve.org/CVERecord?id=CVE-2025-32906
- CVE-2025-32910:
Libsoup: null pointer deference on libsoup via /auth/soup-auth-digest.c
through "soup_auth_digest_authenticate" on client when server omits the
"realm" parameter in an unauthorized response with digest authentication
https://www.cve.org/CVERecord?id=CVE-2025-32910
- CVE-2025-32911:
Libsoup: double free on soup_message_headers_get_content_disposition()
through "soup-message-headers.c" via "params" ghashtable value
https://www.cve.org/CVERecord?id=CVE-2025-32911
- CVE-2025-32912:
Libsoup: null pointer dereference in client when server omits the "nonce"
parameter in an unauthorized response with digest authentication
https://www.cve.org/CVERecord?id=CVE-2025-32912
- CVE-2025-32913:
Libsoup: null pointer dereference in
soup_message_headers_get_content_disposition when "filename" parameter is
present, but has no value in content-disposition header
https://www.cve.org/CVERecord?id=CVE-2025-32913
- CVE-2025-32914:
Libsoup: oob read on libsoup through function
"soup_multipart_new_from_message" in soup-multipart.c leads to crash or
exit of process
https://www.cve.org/CVERecord?id=CVE-2025-32914
- CVE-2025-46420:
Libsoup: memory leak on soup_header_parse_quality_list() via soup-headers.c
https://www.cve.org/CVERecord?id=CVE-2025-46420
- CVE-2025-46421:
Libsoup: information disclosure may leads libsoup client sends authorization
header to a different host when being redirected by a server
https://www.cve.org/CVERecord?id=CVE-2025-46421
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit eee0f6c078)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Commit [1] introduced a patch addressing CVE-2021-23159. Since then,
CVE-2023-34432 ([2]) remained the only unresolved CVE reported
against the sox package in pkg-stat. This patch adds CVE-2023-34432
to the list of ignored CVEs for sox, based on the report from the
Debian Security Tracker ([3]) and the sox issue tracker ([4]), both
indicate that the patch introduced in [1] also resolves this CVE.
[1] 14aa0f5ec1 package/sox: add fix for CVE-2021-23159, CVE-2021-23172, CVE-2023-34318
[2] https://nvd.nist.gov/vuln/detail/CVE-2023-34432
[3] https://security-tracker.debian.org/tracker/CVE-2023-34432
[4] https://sourceforge.net/p/sox/bugs/367/
Signed-off-by: Tim Soubry <tim.soubry@mind.be>
[Julien: change commit ref [1] to use commit id from master branch]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit e868b974a7)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The following error appeared on the autobuilder for host using
host-gcc15:
```
p11_attr.c: In function 'pkcs11_addattr_bool':
p11_attr.c:126:25: error: expected identifier or '(' before 'true'
126 | static CK_BBOOL true = CK_TRUE;
| ^~~~
p11_attr.c:127:25: error: expected identifier or '(' before 'false'
127 | static CK_BBOOL false = CK_FALSE;
| ^~~~~
p11_attr.c:128:44: error: lvalue required as unary '&' operand
128 | pkcs11_addattr(tmpl, type, value ? &true : &false, sizeof(CK_BBOOL));
| ^
p11_attr.c:128:52: error: lvalue required as unary '&' operand
128 | pkcs11_addattr(tmpl, type, value ? &true : &false, sizeof(CK_BBOOL));
| ^
make[3]: *** [Makefile:646: libp11_la-p11_attr.lo] Error 1
```
This is due to the change in the default C language version in GCC15.
This patch backport the upstream patch that fix that issue by not using
the keywords.
Fixes: https://autobuild.buildroot.org/results/da7/da71db9b04f181b9d2e72df73ac8541709f5a1d4
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit df60b105b4)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
This fixes the following vulnerabilities:
- CVE-2023-4256:
Within tcpreplay's tcprewrite, a double free vulnerability has been
identified in the tcpedit_dlt_cleanup() function within
plugins/dlt_plugins.c. This vulnerability can be exploited by
supplying a specifically crafted file to the tcprewrite binary. This
flaw enables a local attacker to initiate a Denial of Service (DoS)
attack.
https://www.cve.org/CVERecord?id=CVE-2023-4256
- CVE-2023-43279:
Null Pointer Dereference in mask_cidr6 component at cidr.c in
Tcpreplay 4.4.4 allows attackers to crash the application via crafted
tcprewrite command.
https://www.cve.org/CVERecord?id=CVE-2023-43279
- CVE-2024-22654:
tcpreplay v4.4.4 was discovered to contain an infinite loop via the
tcprewrite function at get.c.
https://www.cve.org/CVERecord?id=CVE-2024-22654
See the release notes:
https://github.com/appneta/tcpreplay/releases/tag/v4.5.1
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 31619696b9)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
For release note, see:
https://www.sudo.ws/releases/stable/#1.9.17p1
Fixes the following security issues:
- CVE-2025-32462: Sudo before 1.9.17p1, when used with a sudoers file that
specifies a host that is neither the current host nor ALL, allows listed
users to execute commands on unintended machines (since sudo 1.8.8)
https://www.sudo.ws/security/advisories/host_any/
- CVE-2025-32463: Sudo before 1.9.17p1 allows local users to obtain root
access because /etc/nsswitch.conf from a user-controlled directory is used
with the --chroot option (since sudo 1.9.4)
https://www.sudo.ws/security/advisories/chroot_bug/
Update the LICENSE.md hash for a change in copyright years:
30729312c2
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
[Julien: add link to release note in commit log]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit ee86844e63)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The daq package fails to build with GCC14 toolchains:
```
daq_nfq.c: In function 'SetPktHdr':
daq_nfq.c:395:37: error: passing argument 2 of 'nfq_get_payload' from incompatible pointer type [-Wincompatible-pointer-types]
395 | int len = nfq_get_payload(nfad, (char**)pkt);
| ^~~~~~~~~~~
| |
| char **
```
The issue can be reproduced with the following config:
```
cat > daq.config <<EOF
BR2_arm=y
BR2_cortex_a7=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN=y
BR2_PACKAGE_DAQ=y
BR2_PACKAGE_LIBDNET=y
BR2_PACKAGE_LIBNETFILTER_QUEUE=y
EOF
```
This patch port a patch taken from openembedded meta-networking [1] and
is tested with test-pkg:
```
$ ./utils/test-pkg -c daq.config -p daq
```
[1] https://layers.openembedded.org/layerindex/recipe/37594/
Fixes: https://autobuild.buildroot.org/results/c69/c69ab134463a18eec65ded836aecf89a5cb4a75c/build-end.log
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 533c0aac28)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Commit [1] backported an upstream patch to address CVE-2025-46836 that
included a regression.
Upstream later fixed this regression in commit [2].
This patch add that fix to correct the issue introduced by the original
patch.
[1] 323aaa9f54 package/net-tools: add upstream security fix for CVE-2025-46836
[2] ddb0e375fb/
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit d3274210f9)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
For release note, see:
https://github.com/urllib3/urllib3/releases/tag/2.5.0
This fixes the following vulnerabilities:
- CVE-2025-50181:
urllib3 redirects are not disabled when retries are disabled on
PoolManager instantiation
- CVE-2025-50182:
urllib3 does not control redirects in browsers and Node.js
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
[Julien: add link to release note in commit log]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 7006854ce1)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
This fixes the following vulnerability:
- CVE-2023-34194:
StringEqual in TiXmlDeclaration::Parse in tinyxmlparser.cpp in TinyXML
through 2.6.2 has a reachable assertion (and application exit) via a
crafted XML document with a '\0' located after whitespace.
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 184a1b94a5)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Jose-13 fixed the following security issue:
- CVE-2023-50967: latchset jose through version 11 allows attackers to cause
a denial of service (CPU consumption) via a large p2c (aka PBES2 Count)
value.
https://github.com/latchset/jose/issues/151
In addition, jose-14 worked around another DoS issue related to
decompression:
https://github.com/latchset/jose/pull/157
Drop now upstreamed patches:
- 0001-lib-hsh.c-rename-hsh-local-variable.patch: Upstream as of
3d5b287243
- 0002-man-add-option-to-skip-building-man-pages.patch: Upstream after
getting reworked to use -Ddocs=disabled as of
786b426df0
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
[Julien: remove .checkpackageignore entries to fix check-package errors]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 394a8fb406)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The syslinux runtime test (which is in fact a build-only test) ensures
that syslinux does get build at least once a week (via the gitlab-CI
weekly pipeline). Runtime testing would need much more work, though, but
nothing in syslinux is currently runtime tested anyway.
Reported-by: Julien Olivain <ju.o@free.fr>
Signed-off-by: Yann E. MORIN <yann.morin@orange.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit a6ddf2b91d)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The following error occurs on the autobuilder for builds with musl
libc.
```
CC util/bitmap.o
In file included from include/linux/bitmap.h:7,
from util/bitmap.c:9:
include/linux/bitops.h:4:10: fatal error: bits/wordsize.h: No such file or directory
4 | #include <bits/wordsize.h>
| ^~~~~~~~~~~~~~~~~
```
The error occurs because bits/wordsize.h is specific to glibc.
This patch applies an upstream fix that replaces the use of __WORDSIZE
with an internal macro, making the code portable across different libc.
Fixes: https://autobuild.buildroot.org/results/30d/30d6e407e6a0fc7d85062c2d56008755c70ca733/build-end.log
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 901b9e19ed)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The 0.192 release of elfutils introduced the src/srcfiles.cxx program,
that lists all source files of a given ELF binary. As this is a C++
program, we need a toolchain that supports it.
Without it, the build system tries to use "no" as the CXX compiler,
resulting in the following errors :
/bin/sh: line 1: no: command not found
as can be seen here for example :
https://autobuild.buildroot.net/results/849/849221c794a469a423857a290db775d150b84900
Add a dependency to a CPP toolchain for the elfutils programs.
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 268d7ad180)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The elfutils programs require Glibc to be used as the C library. Show a
comment when this libc isn't used in the toolchain.
Suggested-by: Yann E. MORIN <yann.morin@orange.com>
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 52ba3ed657)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
As specified in the 2.28.10 release notes:
https://github.com/Mbed-TLS/mbedtls/releases/tag/mbedtls-2.28.10
Mbed TLS 2.28.10 is the last release of the 2.28 LTS and won't receive bug
fixes or security fixes anymore. Users are advised to upgrade to a
maintained version.
So move to 3.6.x, which is the new LTS version:
Mbed TLS 3.6 is a long-term support (LTS) branch. It will be supported with
bug-fixes and security fixes until at least March 2027.
Drop BR2_PACKAGE_MBEDTLS_COMPRESSION and all related references
as native zlib support has been entirely removed from mbedtls.
Signed-off-by: James Hilliard <james.hilliard1@gmail.com>
[Peter: add note about 2.28.x / 3.6.x, add Config.in.legacy]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit 3481a9643f)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
For release note, see:
https://dev.gnupg.org/T7166
This version fixes a build error which can happen with 32-bit arm
configurations.
The issue can be reproduced with commands:
cat >.config <<EOF
BR2_arm=y
BR2_cortex_a8=y
BR2_ARM_INSTRUCTIONS_THUMB2=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_PACKAGE_LIBGCRYPT=y
EOF
make olddefconfig
make libgcrypt
Build is failing with output:
ec-nist.c: In function '_gcry_mpi_ec_nist256_mod':
ec-inline.h:902:5: error: 'asm' operand has impossible constraints or there are not enough registers
902 | __asm__ ("subs %3, %7, %10\n" \
| ^~~~~~~
Details for this buggix: https://dev.gnupg.org/T7226
Signed-off-by: Bram Oosterhuis <dev@bybram.com>
[Julien: reword commit log]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 89ca1bd4f4)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
syslinux is... special. It is a target package, but it is installed in
HOST_DIR *in the target install commands*: in addition to the boot files
that run on the target, syslinux installs a set of host tools that are
to be used at build time (e.g. extlinux, to prepare bootable media, like
an iso96660 image). Then, from HOST_DIR, the actual boot files are
copied into BINARIES_DIR (i.e. images/); we do it that way because the
boot files are scattered about everywhere in the build tree, while they
are all packed together in a single directory once installed.
However, there is no dependency between the target and image install
steps. So, when using top-level parallel builds, there is no guarantee
that the target install commands are finished before the image install
commands are started.
We fix that by first installing into a temporary location, as part of
the build step, and by then copying from there as part of the install
step. This ensures that the boot files are easily available, without
needing a dependency on the target install step, that we can't express.
Note that we do not change the actual installation into HOST_DIR: it can
be set up differently that our temporary location, and we do not want
to duplicate that setup here (it's going to diverge over time).
Signed-off-by: Yann E. MORIN <yann.morin@orange.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 90e76818a1)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
