diff mbox series

[2/2] package/pistache: enable on uClibc

Message ID 20220804203555.29029-2-thomas.petazzoni@bootlin.com
State Accepted
Headers show
Series [1/2] package/pistache: fix dependency on rapidjson | expand

Commit Message

Thomas Petazzoni Aug. 4, 2022, 8:35 p.m. UTC 
pistache couldn't build on uClibc due to just one missing
definition. In addition, the Config.in comment was wrong as it said
that a glibc toolchain was needed, while in fact it could be glibc or
musl.

Cc: Thomas Ruschival <thomas@ruschival.de>
Cc: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
---
 ...port.cc-fallback-value-for-RUSAGE_TH.patch | 33 +++++++++++++++++++
 package/pistache/Config.in                    |  4 +--
 2 files changed, 34 insertions(+), 3 deletions(-)
 create mode 100644 package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch

Comments

Yann E. MORIN Aug. 5, 2022, 4:29 p.m. UTC | #1 
Thomas², All,

Thomas R., a note for you toward the end...

On 2022-08-04 22:35 +0200, Thomas Petazzoni via buildroot spake thusly:
> pistache couldn't build on uClibc due to just one missing
> definition. In addition, the Config.in comment was wrong as it said
> that a glibc toolchain was needed, while in fact it could be glibc or
> musl.

As I understand, this patch is not really upstreamable, or at least not
in this state.

However, this means we will have to carry over and over, because
existing uclibc-based toolchains will still be usad in the future, even
when newer uclibc-ng contains the new definitions, so we can't drop this
patch because it may always needed when using external toolchains...

So, I think a version that is upstreamable would be nice...

Thomas R.: Thomas P. noticed a very nasty beha  viour for pistache: it
will forcibly register git hooks when it is built, if it finds that it
is built in a git tree.

This is very nasty, because in our case, developping Buildroot the
pistache sources are extacted from a tarball, but there is a git tree to
be found higher up: it finds Buildroot's own git tree, s registers git
hooks in that repository, and the results is obviously wrong and
prevents properly working:

    thomas@windsurf:~/projets/buildroot (pistache-fix)$ make pistache-extract
    pistache-3ec9d7c4f8b828fdd391550fff81b01e72dd6269.tar.gz: OK (sha256: f2b3e8b2581cfed5e036d1b97a9b97dc7022b3ddaab69c4691238faff3199bc1)
    >>> pistache 3ec9d7c4f8b828fdd391550fff81b01e72dd6269 Extracting
    gzip -d -c /home/thomas/dl/pistache/pistache-3ec9d7c4f8b828fdd391550fff81b01e72dd6269.tar.gz | tar --strip-components=1 -C /home/thomas/projets/buildroot/output/build/pistache-3ec9d7c4f8b828fdd391550fff81b01e72dd6269   -xf -

    thomas@windsurf:~/projets/buildroot (pistache-fix)$ git commit
    grep: version.txt: No such file or directory
    Update the commit date in version.txt before committing

    thomas@windsurf:~/projets/buildroot (pistache-fix)$ grep pistache .git/config
            hooksPath = /home/thomas/projets/buildroot/output/build/pistache-3ec9d7c4f8b828fdd391550fff81b01e72dd6269/.hooks

See: https://github.com/pistacheio/pistache/blob/master/meson.build#L137 :

    if not meson.is_subproject()
        git = find_program('git', required: false)
        if git.found()
            run_command(git, 'config', '--local', 'core.hooksPath', meson.source_root()/'.hooks', check: false)
        endif
    endif

Could you look into proposing a change upstream, that at the very least
makes that hook registration optional?

Regards,
Yann E. MORIN.

> Cc: Thomas Ruschival <thomas@ruschival.de>
> Cc: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
> ---
>  ...port.cc-fallback-value-for-RUSAGE_TH.patch | 33 +++++++++++++++++++
>  package/pistache/Config.in                    |  4 +--
>  2 files changed, 34 insertions(+), 3 deletions(-)
>  create mode 100644 package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch
> 
> diff --git a/package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch b/package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch
> new file mode 100644
> index 0000000000..327933e26d
> --- /dev/null
> +++ b/package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch
> @@ -0,0 +1,33 @@
> +From 5feed6d2f86750bfe1e7c26d6fd4e8d2195ea81f Mon Sep 17 00:00:00 2001
> +From: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
> +Date: Thu, 4 Aug 2022 19:54:23 +0200
> +Subject: [PATCH] src/common/transport.cc: fallback value for RUSAGE_THREAD
> +
> +uClibc-ng currently does not define RUSAGE_THREAD, so let's add a
> +fallback value until it gets fixed in uClibc-ng. The value is the same
> +on all Linux systems, regardless of the CPU architecture.
> +
> +Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
> +Upstream bug: https://mailman.openadk.org/mailman3/hyperkitty/list/devel@uclibc-ng.org/thread/MEXOUSFSH46UFS3AEMEIEPTL5U4HVEFL/
> +---
> + src/common/transport.cc | 4 ++++
> + 1 file changed, 4 insertions(+)
> +
> +diff --git a/src/common/transport.cc b/src/common/transport.cc
> +index 62efb8c..0f3abb0 100644
> +--- a/src/common/transport.cc
> ++++ b/src/common/transport.cc
> +@@ -515,6 +515,10 @@ namespace Pistache::Tcp
> + 
> +         rusage now;
> + 
> ++#if !defined(RUSAGE_THREAD)
> ++#define RUSAGE_THREAD 1
> ++#endif
> ++
> +         auto res = getrusage(RUSAGE_THREAD, &now);
> +         if (res == -1)
> +             loadRequest_.reject(std::runtime_error("Could not compute usage"));
> +-- 
> +2.37.1
> +
> diff --git a/package/pistache/Config.in b/package/pistache/Config.in
> index 53a4b73a58..3acf1f53f0 100644
> --- a/package/pistache/Config.in
> +++ b/package/pistache/Config.in
> @@ -4,7 +4,6 @@ config BR2_PACKAGE_PISTACHE
>  	depends on BR2_USE_WCHAR
>  	depends on BR2_TOOLCHAIN_HAS_THREADS
>  	depends on BR2_INSTALL_LIBSTDCPP
> -	depends on !BR2_TOOLCHAIN_USES_UCLIBC
>  	depends on !BR2_TOOLCHAIN_HAS_BINUTILS_BUG_27597
>  	select BR2_PACKAGE_RAPIDJSON
>  	help
> @@ -14,9 +13,8 @@ config BR2_PACKAGE_PISTACHE
>  
>  	  https://github.com/oktal/pistache
>  
> -comment "pistache needs a glibc toolchain w/ C++, gcc >= 7, threads, wchar, not binutils bug 27597"
> +comment "pistache needs a toolchain w/ C++, gcc >= 7, threads, wchar, not binutils bug 27597"
>  	depends on !BR2_INSTALL_LIBSTDCPP || \
> -		   BR2_TOOLCHAIN_USES_UCLIBC || \
>  		   !BR2_TOOLCHAIN_GCC_AT_LEAST_7 || \
>  		   !BR2_TOOLCHAIN_HAS_THREADS || \
>  		   !BR2_USE_WCHAR || \
> -- 
> 2.37.1
> 
> _______________________________________________
> buildroot mailing list
> buildroot@buildroot.org
> https://lists.buildroot.org/mailman/listinfo/buildroot
Thomas Petazzoni Aug. 6, 2022, 1:45 p.m. UTC | #2 
Hello,

On Fri, 5 Aug 2022 18:29:06 +0200
"Yann E. MORIN" <yann.morin.1998@free.fr> wrote:

> As I understand, this patch is not really upstreamable, or at least not
> in this state.
> 
> However, this means we will have to carry over and over, because
> existing uclibc-based toolchains will still be usad in the future, even
> when newer uclibc-ng contains the new definitions, so we can't drop this
> patch because it may always needed when using external toolchains...

No, we will not have to carry it over and over. Regarding uClibc-ng, we
require using a reasonably recent version of uClibc-ng. So my plan
would be to keep this hack for a few months, until at least the Bootlin
uClibc toolchains are rebuilt with a recent enough uClibc-ng, and then
drop this hack.

So no, the goal is not to keep this hack patch forever.

What I'm proposing to do for pistache is exactly the same as I did in
commit fa4dba4aa202de20e3821eace5513aad81c23275, for openswan. Waldemar
was simply proposing to re-enable openswan for uClibc, because
uClibc-ng 1.0.41 has the necessary definitions. But in the immediate
term, that would have caused build break with Bootlin toolchains. So
until they are re-generated, I have added a hack patch that avoids the
uClibc build issue entirely, also with the goal of dropping the patch
when the toolchains are rebuilt.

> So, I think a version that is upstreamable would be nice...

I'm not sure it's worth the effort to introduce more complexity
upstream for this missing definition in uClibc. We should just fix it
in uClibc (I have already reported the issue), and that's it.

Best regards,

Thomas
Yann E. MORIN Aug. 6, 2022, 2:15 p.m. UTC | #3 
Thomas, All,

On 2022-08-04 22:35 +0200, Thomas Petazzoni via buildroot spake thusly:
> pistache couldn't build on uClibc due to just one missing
> definition. In addition, the Config.in comment was wrong as it said
> that a glibc toolchain was needed, while in fact it could be glibc or
> musl.
> 
> Cc: Thomas Ruschival <thomas@ruschival.de>
> Cc: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

Applied to master, thanks.

Regards,
Yann E. MORIN.

> ---
>  ...port.cc-fallback-value-for-RUSAGE_TH.patch | 33 +++++++++++++++++++
>  package/pistache/Config.in                    |  4 +--
>  2 files changed, 34 insertions(+), 3 deletions(-)
>  create mode 100644 package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch
> 
> diff --git a/package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch b/package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch
> new file mode 100644
> index 0000000000..327933e26d
> --- /dev/null
> +++ b/package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch
> @@ -0,0 +1,33 @@
> +From 5feed6d2f86750bfe1e7c26d6fd4e8d2195ea81f Mon Sep 17 00:00:00 2001
> +From: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
> +Date: Thu, 4 Aug 2022 19:54:23 +0200
> +Subject: [PATCH] src/common/transport.cc: fallback value for RUSAGE_THREAD
> +
> +uClibc-ng currently does not define RUSAGE_THREAD, so let's add a
> +fallback value until it gets fixed in uClibc-ng. The value is the same
> +on all Linux systems, regardless of the CPU architecture.
> +
> +Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
> +Upstream bug: https://mailman.openadk.org/mailman3/hyperkitty/list/devel@uclibc-ng.org/thread/MEXOUSFSH46UFS3AEMEIEPTL5U4HVEFL/
> +---
> + src/common/transport.cc | 4 ++++
> + 1 file changed, 4 insertions(+)
> +
> +diff --git a/src/common/transport.cc b/src/common/transport.cc
> +index 62efb8c..0f3abb0 100644
> +--- a/src/common/transport.cc
> ++++ b/src/common/transport.cc
> +@@ -515,6 +515,10 @@ namespace Pistache::Tcp
> + 
> +         rusage now;
> + 
> ++#if !defined(RUSAGE_THREAD)
> ++#define RUSAGE_THREAD 1
> ++#endif
> ++
> +         auto res = getrusage(RUSAGE_THREAD, &now);
> +         if (res == -1)
> +             loadRequest_.reject(std::runtime_error("Could not compute usage"));
> +-- 
> +2.37.1
> +
> diff --git a/package/pistache/Config.in b/package/pistache/Config.in
> index 53a4b73a58..3acf1f53f0 100644
> --- a/package/pistache/Config.in
> +++ b/package/pistache/Config.in
> @@ -4,7 +4,6 @@ config BR2_PACKAGE_PISTACHE
>  	depends on BR2_USE_WCHAR
>  	depends on BR2_TOOLCHAIN_HAS_THREADS
>  	depends on BR2_INSTALL_LIBSTDCPP
> -	depends on !BR2_TOOLCHAIN_USES_UCLIBC
>  	depends on !BR2_TOOLCHAIN_HAS_BINUTILS_BUG_27597
>  	select BR2_PACKAGE_RAPIDJSON
>  	help
> @@ -14,9 +13,8 @@ config BR2_PACKAGE_PISTACHE
>  
>  	  https://github.com/oktal/pistache
>  
> -comment "pistache needs a glibc toolchain w/ C++, gcc >= 7, threads, wchar, not binutils bug 27597"
> +comment "pistache needs a toolchain w/ C++, gcc >= 7, threads, wchar, not binutils bug 27597"
>  	depends on !BR2_INSTALL_LIBSTDCPP || \
> -		   BR2_TOOLCHAIN_USES_UCLIBC || \
>  		   !BR2_TOOLCHAIN_GCC_AT_LEAST_7 || \
>  		   !BR2_TOOLCHAIN_HAS_THREADS || \
>  		   !BR2_USE_WCHAR || \
> -- 
> 2.37.1
> 
> _______________________________________________
> buildroot mailing list
> buildroot@buildroot.org
> https://lists.buildroot.org/mailman/listinfo/buildroot
diff mbox series

Patch

diff --git a/package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch b/package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch
new file mode 100644
index 0000000000..327933e26d
--- /dev/null
+++ b/package/pistache/0001-src-common-transport.cc-fallback-value-for-RUSAGE_TH.patch
@@ -0,0 +1,33 @@ 
+From 5feed6d2f86750bfe1e7c26d6fd4e8d2195ea81f Mon Sep 17 00:00:00 2001
+From: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
+Date: Thu, 4 Aug 2022 19:54:23 +0200
+Subject: [PATCH] src/common/transport.cc: fallback value for RUSAGE_THREAD
+
+uClibc-ng currently does not define RUSAGE_THREAD, so let's add a
+fallback value until it gets fixed in uClibc-ng. The value is the same
+on all Linux systems, regardless of the CPU architecture.
+
+Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
+Upstream bug: https://mailman.openadk.org/mailman3/hyperkitty/list/devel@uclibc-ng.org/thread/MEXOUSFSH46UFS3AEMEIEPTL5U4HVEFL/
+---
+ src/common/transport.cc | 4 ++++
+ 1 file changed, 4 insertions(+)
+
+diff --git a/src/common/transport.cc b/src/common/transport.cc
+index 62efb8c..0f3abb0 100644
+--- a/src/common/transport.cc
++++ b/src/common/transport.cc
+@@ -515,6 +515,10 @@ namespace Pistache::Tcp
+ 
+         rusage now;
+ 
++#if !defined(RUSAGE_THREAD)
++#define RUSAGE_THREAD 1
++#endif
++
+         auto res = getrusage(RUSAGE_THREAD, &now);
+         if (res == -1)
+             loadRequest_.reject(std::runtime_error("Could not compute usage"));
+-- 
+2.37.1
+
diff --git a/package/pistache/Config.in b/package/pistache/Config.in
index 53a4b73a58..3acf1f53f0 100644
--- a/package/pistache/Config.in
+++ b/package/pistache/Config.in
@@ -4,7 +4,6 @@  config BR2_PACKAGE_PISTACHE
 	depends on BR2_USE_WCHAR
 	depends on BR2_TOOLCHAIN_HAS_THREADS
 	depends on BR2_INSTALL_LIBSTDCPP
-	depends on !BR2_TOOLCHAIN_USES_UCLIBC
 	depends on !BR2_TOOLCHAIN_HAS_BINUTILS_BUG_27597
 	select BR2_PACKAGE_RAPIDJSON
 	help
@@ -14,9 +13,8 @@  config BR2_PACKAGE_PISTACHE
 
 	  https://github.com/oktal/pistache
 
-comment "pistache needs a glibc toolchain w/ C++, gcc >= 7, threads, wchar, not binutils bug 27597"
+comment "pistache needs a toolchain w/ C++, gcc >= 7, threads, wchar, not binutils bug 27597"
 	depends on !BR2_INSTALL_LIBSTDCPP || \
-		   BR2_TOOLCHAIN_USES_UCLIBC || \
 		   !BR2_TOOLCHAIN_GCC_AT_LEAST_7 || \
 		   !BR2_TOOLCHAIN_HAS_THREADS || \
 		   !BR2_USE_WCHAR || \