Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

classic Classic list List threaded Threaded
13 messages Options
Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Santiago Vila-2
Package: src:webkit2gtk
Version: 2.22.0-2
Tags: ftbfs

Dear maintainer:

I tried to build this package in buster but it failed:

--------------------------------------------------------------------------------
[...]
 debian/rules build-indep
dh build-indep --with gir
   dh_update_autotools_config -i
   dh_autoreconf -i
   debian/rules override_dh_auto_configure
make[1]: Entering directory '/<<PKGBUILDDIR>>'
CXXFLAGS="-g -O2 -fdebug-prefix-map=/<<PKGBUILDDIR>>=. -fstack-protector-strong -Wformat -Werror=format-security" \
dh_auto_configure -- \
   -DPORT=GTK \
   -DCMAKE_INSTALL_LIBEXECDIR=lib/x86_64-linux-gnu \
   -DCMAKE_C_FLAGS_RELEASE="" \

[... snipped ...]

html/webkitdomgtk-4.0/index-all.html:5246: warning: no link for: "WebKitDOMClientRectList-struct" -> (WebKitDOMClientRectList).
html/webkitdomgtk-4.0/index-all.html:5250: warning: no link for: "WebKitDOMClientRectList--length" -> (WebKitDOMClientRectList:length).
html/webkitdomgtk-4.0/webkitdomgtk-4.0-WebKitDOMHTMLBaseFontElement.html:102: warning: no link for: "WebKitDOMHTMLBaseFontElement" -> (<span class="type">WebKitDOMHTMLBaseFontElement</span>).
html/webkitdomgtk-4.0/webkitdomgtk-4.0-WebKitDOMNodeFilter.html:180: warning: no link for: "api-index-2.6" -> (2.6).
html/webkitdomgtk-4.0/webkitdomgtk-4.0-WebKitDOMNodeFilter.html:229: warning: no link for: "WebKitDOMCDataSection" -> (<span class="type">WebKitDOMCDataSection</span>).
html/webkitdomgtk-4.0/webkitdomgtk-4.0-WebKitDOMNodeFilter.html:301: warning: no link for: "WebKitDOMEntity" -> (<span class="type">WebKitDOMEntity</span>).
html/webkitdomgtk-4.0/webkitdomgtk-4.0-WebKitDOMNodeFilter.html:325: warning: no link for: "WebKitDOMNotation" -> (<span class="type">WebKitDOMNotation</span>).
html/webkit2gtk-4.0/WebKitSettings.html:2617: warning: no link for: "WebKitWebContext--is-ephemeral" -> (<span class="type">“is-ephemeral”</span>).
html/webkit2gtk-4.0/WebKitURIResponse.html:195: warning: no link for: "SoupKnownStatusCode" -> (<span class="type">SoupKnownStatusCode</span>).
html/webkit2gtk-4.0/WebKitWebExtension.html:112: warning: no link for: "webkit-web-extension-initialize-with-user-data" -> (<code class="function">webkit_web_extension_initialize_with_user_data()</code>).
html/webkit2gtk-4.0/WebKitWebInspector.html:248: warning: no link for: "WebKitWebViewBase" -> (<span class="type">WebKitWebViewBase</span>).
html/webkit2gtk-4.0/WebKitWebPage.html:673: warning: no link for: "WebKitFormSubmissionEventType" -> (<span class="type">WebKitFormSubmissionEventType</span>).
touch docs-build.stamp
make[3]: Leaving directory '/<<PKGBUILDDIR>>/obj-x86_64-linux-gnu'
[ 92%] Built target gtkdoc
make -f Source/WebKit/CMakeFiles/WebKit2WebExtension-4-gir.dir/build.make Source/WebKit/CMakeFiles/WebKit2WebExtension-4-gir.dir/depend
make[3]: Entering directory '/<<PKGBUILDDIR>>/obj-x86_64-linux-gnu'
cd /<<PKGBUILDDIR>>/obj-x86_64-linux-gnu && /usr/bin/cmake -E cmake_depends "Unix Makefiles" /<<PKGBUILDDIR>> /<<PKGBUILDDIR>>/Source/WebKit /<<PKGBUILDDIR>>/obj-x86_64-linux-gnu /<<PKGBUILDDIR>>/obj-x86_64-linux-gnu/Source/WebKit /<<PKGBUILDDIR>>/obj-x86_64-linux-gnu/Source/WebKit/CMakeFiles/WebKit2WebExtension-4-gir.dir/DependInfo.cmake --color=
Scanning dependencies of target WebKit2WebExtension-4-gir
make[3]: Leaving directory '/<<PKGBUILDDIR>>/obj-x86_64-linux-gnu'
make -f Source/WebKit/CMakeFiles/WebKit2WebExtension-4-gir.dir/build.make Source/WebKit/CMakeFiles/WebKit2WebExtension-4-gir.dir/build
make[3]: Entering directory '/<<PKGBUILDDIR>>/obj-x86_64-linux-gnu'
make[3]: *** No rule to make target 'JavaScriptCore-4.0.gir', needed by 'WebKit2-4.0.gir'.  Stop.
make[3]: Leaving directory '/<<PKGBUILDDIR>>/obj-x86_64-linux-gnu'
make[2]: *** [CMakeFiles/Makefile2:1605: Source/WebKit/CMakeFiles/WebKit2WebExtension-4-gir.dir/all] Error 2
make[2]: Leaving directory '/<<PKGBUILDDIR>>/obj-x86_64-linux-gnu'
make[1]: *** [Makefile:155: all] Error 2
make[1]: Leaving directory '/<<PKGBUILDDIR>>/obj-x86_64-linux-gnu'
dh_auto_build: cd obj-x86_64-linux-gnu && make -j1 returned exit code 2
make: *** [debian/rules:57: build-indep] Error 2
dpkg-buildpackage: error: debian/rules build-indep subprocess returned exit status 2
--------------------------------------------------------------------------------

I'm using sbuild + schroot + eatmydata, but the part saying

No rule to make target 'JavaScriptCore-4.0.gir', needed by 'WebKit2-4.0.gir'

suggests a Makefile bug.

I could probably prepare a VM instance in which this happens (please contact me
privately for details), but I believe a close look at the Makefile should perhaps
tell what's wrong without having to "reproduce" it.

In either case, I've put a bunch of failed build logs here for you to see:

https://people.debian.org/~sanvila/build-logs/webkit2gtk/

Thanks.

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Alberto Garcia-2
On Thu, Feb 28, 2019 at 05:19:40PM +0000, Santiago Vila wrote:
> make[3]: *** No rule to make target 'JavaScriptCore-4.0.gir', needed by 'WebKit2-4.0.gir'.  Stop.

I think I've come across this error a couple of times, does it go away
if you try to rebuild?

We are considering to use --buildsystem=cmake+ninja for the build,
perhaps that solves this problem (I haven't tried it yet).

> In either case, I've put a bunch of failed build logs here for you to see:
>
> https://people.debian.org/~sanvila/build-logs/webkit2gtk/

Thanks. Does it fail all the time then?

Berto

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Santiago Vila
> > In either case, I've put a bunch of failed build logs here for you to see:
> >
> > https://people.debian.org/~sanvila/build-logs/webkit2gtk/
>
> Thanks. Does it fail all the time then?

In my environment, yes. It failed every time I tried with each new
Debian release since 2.22.0-2. Here is my build history:

Status: successful  webkit2gtk_2.18.5-1_amd64-20180118T052922Z
Status: successful  webkit2gtk_2.18.6-1_amd64-20180128T002903Z
Status: successful  webkit2gtk_2.20.4-1_amd64-20180809T071009Z
Status: successful  webkit2gtk_2.20.5-1_amd64-20180825T104210Z
Status: failed      webkit2gtk_2.22.0-2_amd64-20180920T034119Z
Status: failed      webkit2gtk_2.22.2-1_amd64-20180927T034115Z
Status: failed      webkit2gtk_2.22.3-1_amd64-20181103T035412Z
Status: failed      webkit2gtk_2.22.4-1_amd64-20181128T091510Z
Status: failed      webkit2gtk_2.22.5-1_amd64-20181219T230911.565Z
Status: failed      webkit2gtk_2.22.6-1_amd64-20190212T225410.081Z

For this package, my environment is made of self-hosted KVM machines
with a single CPU. Maybe having a single CPU makes the bug
particularly easy to trigger, but being a Makefile bug, it could
happen to anybody.

Please tell me if you need a test machine and I'll try to prepare one,
but as I said before, maybe it would be more productive to try to
explain how the error I got could happen (for example: It is true, as
the error says, that there is not any rule to build
JavaScriptCore-4.0.gir?)

Thanks.

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Jeremy Bicha-5
On Fri, Mar 1, 2019 at 4:36 AM Santiago Vila <[hidden email]> wrote:
> For this package, my environment is made of self-hosted KVM machines
> with a single CPU. Maybe having a single CPU makes the bug
> particularly easy to trigger, but being a Makefile bug, it could
> happen to anybody.

Honestly, you might be the only person to want to try to build
webkitgtk with a single CPU core.

Why don't you try cmake+ninja? See
https://salsa.debian.org/webkit-team/webkit/merge_requests/3

Thanks,
Jeremy Bicha

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Alberto Garcia-2
In reply to this post by Santiago Vila
Control: forwarded -1 https://bugs.webkit.org/show_bug.cgi?id=195251

On Fri, Mar 01, 2019 at 10:32:40AM +0100, Santiago Vila wrote:

> > Thanks. Does it fail all the time then?
>
> In my environment, yes. It failed every time I tried with each new
> Debian release since 2.22.0-2.

I confirm that it fails with make -j1, it seems that there's indeed no
rule to make JavaScriptCore-4.0.gir.

'make JavaScriptCore-4-gir' does work, so I suspect that there's a
missing/wrong rule in the CMake files.

Berto

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Santiago Vila
In reply to this post by Jeremy Bicha-5
On Sat, Mar 02, 2019 at 08:29:05AM -0500, Jeremy Bicha wrote:
> On Fri, Mar 1, 2019 at 4:36 AM Santiago Vila <[hidden email]> wrote:
> > For this package, my environment is made of self-hosted KVM machines
> > with a single CPU. Maybe having a single CPU makes the bug
> > particularly easy to trigger, but being a Makefile bug, it could
> > happen to anybody.
>
> Honestly, you might be the only person to want to try to build
> webkitgtk with a single CPU core.

I find your CPUism disturbing. This is a bug in the Makefile, not a bug
in my building environment.

Period.

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Santiago Vila
In reply to this post by Alberto Garcia-2
Hi. I built version 2.23.92-1 in experimental in four different
autobuilders and it worked ok the four times:

https://people.debian.org/~sanvila/build-logs/webkit2gtk/

So even if we don't really understand at this moment the reason it
failed before, I would consider ok to do an upload for unstable
(having an equivalent fix) and close the bug.

Thanks.

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Alberto Garcia-2
On Fri, Mar 08, 2019 at 12:42:48AM +0100, Santiago Vila wrote:
> Hi. I built version 2.23.92-1 in experimental in four different
> autobuilders and it worked ok the four times:
>
> https://people.debian.org/~sanvila/build-logs/webkit2gtk/

Thanks for testing it!

Possible reasons:

   - There's actually not a bug on the cmake files themselves, but a
     bug on the cmake-to-Makefile generator that is not present in the
     camke-to-ninja generator. I don't think that's the case, but it
     could perfectly be.

   - The cmake-to-ninja generator (or ninja itself) happens to build
     the target in a different order than make, one that doesn't
     trigger the problem.

Either way I would let the upstream bug open until someone confirms
where the problem is.

For Debian the change is very simple (apply the patch to use
cmake+ninja to the unstable branch), are we still on time to upload
this? In any case I'd wait until 2.22.7-1 reaches buster first, it has
important fixes.

Berto

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Jeremy Bicha-5
On Fri, Mar 8, 2019 at 4:03 AM Alberto Garcia <[hidden email]> wrote:
> For Debian the change is very simple (apply the patch to use
> cmake+ninja to the unstable branch), are we still on time to upload
> this? In any case I'd wait until 2.22.7-1 reaches buster first, it has
> important fixes.

Since 2.22 is unsupportable for Buster (upstream will stop supporting
it this month), we'll be filing an unblock request for 2.24 after it's
released next week.

2.22.7-1 might need an unblock too because it might be just a bit too
late for the hard code freeze.

Thanks,
Jeremy Bicha

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Santiago Vila
On Fri, Mar 08, 2019 at 05:40:01AM -0500, Jeremy Bicha wrote:
> On Fri, Mar 8, 2019 at 4:03 AM Alberto Garcia <[hidden email]> wrote:
> > For Debian the change is very simple (apply the patch to use
> > cmake+ninja to the unstable branch), are we still on time to upload
> > this? In any case I'd wait until 2.22.7-1 reaches buster first, it has
> > important fixes.
>
> Since 2.22 is unsupportable for Buster (upstream will stop supporting
> it this month), we'll be filing an unblock request for 2.24 after it's
> released next week.

So: Can we expect to see 2.24 in buster, given the unsupportable
status of 2.22? (If you need to talk with RM, please do so).

Thanks.

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS for unknown reasons (maybe Makefile bug)

Alberto Garcia-2
On Mon, Apr 08, 2019 at 06:02:41PM +0200, Santiago Vila wrote:

> On Fri, Mar 08, 2019 at 05:40:01AM -0500, Jeremy Bicha wrote:
> > On Fri, Mar 8, 2019 at 4:03 AM Alberto Garcia <[hidden email]> wrote:
> > > For Debian the change is very simple (apply the patch to use
> > > cmake+ninja to the unstable branch), are we still on time to upload
> > > this? In any case I'd wait until 2.22.7-1 reaches buster first, it has
> > > important fixes.
> >
> > Since 2.22 is unsupportable for Buster (upstream will stop supporting
> > it this month), we'll be filing an unblock request for 2.24 after it's
> > released next week.
>
> So: Can we expect to see 2.24 in buster, given the unsupportable
> status of 2.22? (If you need to talk with RM, please do so).

I plan to upload it to unstable this week (2.24.1 is about to be
released).

Berto

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS when built with parallel=1

Santiago Vila
close 923476 2.24.1-1
retitle 923476 webkit2gtk: FTBFS when built with parallel=1 (No rule to make target 'JavaScriptCore-4.0.gir', needed by 'WebKit2-4.0.gir')
thanks

(Note: The retitle merely adds a snippet from the build log to the title set by Alberto)

This works ok in the single-cpu systems I've tested, therefore I'm
closing this with the version in unstable.

Can we already raise this to serious, where it probably belongs?
I think it would also help for the unstable -> testing transition.

Thanks.

Reply | Threaded
Open this post in threaded view
|

Bug#923476: webkit2gtk: FTBFS when built with parallel=1

Alberto Garcia-2
On Wed, Apr 10, 2019 at 09:22:33AM +0200, Santiago Vila wrote:
> close 923476 2.24.1-1
> retitle 923476 webkit2gtk: FTBFS when built with parallel=1 (No rule to make target 'JavaScriptCore-4.0.gir', needed by 'WebKit2-4.0.gir')
> thanks
>
> (Note: The retitle merely adds a snippet from the build log to the title set by Alberto)
>
> This works ok in the single-cpu systems I've tested, therefore I'm
> closing this with the version in unstable.

Here's the unblock request:

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=926929

Berto