Alle lunedì 7 dicembre 2009, Cyril Brulebois ha scritto:
> I believe I found the issue: config.h's line 6 is about prefix. As can
> be seen in a cowbuilder build log:
> | -- Found automoc4: /usr/bin/automoc4
> | ** PREFIX=/usr
> | -- Configuring done
> | -- Generating done
> while in a buildd build log:
> | -- Found automoc4: /usr/bin/automoc4
> | ** PREFIX=trying to create local folder /home/buildd: Permission denied
> | /usr
> | -- Configuring done
> | -- Generating done
> So I guess it might be the reason why we're then getting .h files that
> are fucked up, leading to a broken build afterwards.
> Trying to reproduce it on a porter box where the package otherwise
> builds fine, I can indeed trigger this FTBFS:
> | $ HOME=/foobar debuild -eHOME -B

Excellent finding, I have prepared a patch that should make it build even with 
read-only home. Patch (with a bit longer explanation) is attached.
Looks working correctly here, feel free to give it a test.

Pino Toscano
From: Pino Toscano <p...@kde.org>
Description: Disable braindead KDE3/4 prefix detection
 The current detection of the KDE 3 and KDE 4 prefixes is done running
 kde-config and kde4-config, resp.
 This can be problematic, as these two could create files in ~/.kde,
 thus breaking when the home of the current user (eg in buildds) is read-only.
 Given the build system sets anyway "/usr" as prefix of KDE 3 and KDE 4
 if they were not detected, then just let it to.
Bug-Debian: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=541667

--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -16,6 +16,11 @@
+# Comment the kde3/kde4 prefix detection as a whole as it wants to run
+# kde-config and kde4-config, which in turn can create stuff in ~/.kde.
+# This cannot work on read-only homes.
+# So, the "fallback" prefixes ("/usr" for both) will be used.
 # Try to find kde4-config...
@@ -48,6 +53,7 @@
                     ARGS --prefix
                     OUTPUT_VARIABLE KDE4PREFIX)

Attachment: signature.asc
Description: This is a digitally signed message part.

pkg-kde-extras mailing list

Reply via email to