Re: [mpd-devel] [PATCH] Workaround: open() ignores 'mode' with O_TMPFILE

2015-05-21 Thread Max Kellermann
On 2015/05/21 09:02, Yclept Nemo orbisvi...@gmail.com wrote: And how do you know the bug is the same on the build machine as on the machine running MPD? Hmm, I didn't consider this. The problem is not that the build machine suffers the bug while distribution machines don't, but vice

Re: [mpd-devel] [PATCH] Workaround: open() ignores 'mode' with O_TMPFILE

2015-05-21 Thread Yclept Nemo
On Thu, May 21, 2015 at 2:17 AM, Max Kellermann m...@duempel.org wrote: Whose bug is this, where's the bug report? http://mailman.blarg.de/pipermail/mpd-devel/2015-May/000320.html And how do you know the bug is the same on the build machine as on the machine running MPD? Hmm, I didn't

Re: [mpd-devel] [PATCH] Workaround: open() ignores 'mode' with O_TMPFILE

2015-05-21 Thread Yclept Nemo
On Thu, May 21, 2015 at 3:02 AM, Yclept Nemo orbisvi...@gmail.com wrote: On Thu, May 21, 2015 at 2:17 AM, Max Kellermann m...@duempel.org wrote: And how do you know the bug is the same on the build machine as on the machine running MPD? Hmm, I didn't consider this. The problem is not that

Re: [mpd-devel] [PATCH] Workaround: open() ignores 'mode' with O_TMPFILE

2015-05-21 Thread Max Kellermann
On 2015/05/21 09:08, Yclept Nemo orbisvi...@gmail.com wrote: OK, the test case should be moved from configure.ac into the constructor of FileDescriptor. If you do this test every time a FileDescriptor gets constructed, what's the advantage of testing it in the first place?

Re: [mpd-devel] Upstream MPD aborts on second run

2015-05-21 Thread Yclept Nemo
On Wed, May 20, 2015 at 3:43 AM, Yclept Nemo orbisvi...@gmail.com wrote: * This is either related to or identical and fixed by the following bug-report (2015-02-24): * https://sourceware.org/bugzilla/show_bug.cgi?id=17523 Confirmed, it's the same bug. This is fixed in upstream glibc by

Re: [mpd-devel] [PATCH 1/6] InputStream: add ReadFull method

2015-05-21 Thread Max Kellermann
On 2015/05/21 23:43, Thomas Guillem tho...@gllm.fr wrote: Convenient method that behave differently than Read, and that will be used by tag scanners. This method will return in case of error, if the whole data is read or is EOF is reached. But how will the caller know that an error has

[mpd-devel] [PATCH 2/6] tag: Ape/Riff/Aiff/Id3 now use an InputStream

2015-05-21 Thread Thomas Guillem
This allow to use these scanners for remote InputStream like SMB and NFS. --- src/tag/Aiff.cxx | 29 +-- src/tag/Aiff.hxx | 4 +- src/tag/ApeLoader.cxx | 43 +- src/tag/ApeLoader.hxx | 11 ++ src/tag/ApeTag.cxx| 19 ++

[mpd-devel] [PATCH 5/6] EmbeddedCuePlaylist: use one InputStream for both scanners

2015-05-21 Thread Thomas Guillem
This avoid to create and delete two InputStream. --- src/playlist/plugins/EmbeddedCuePlaylistPlugin.cxx | 22 -- 1 file changed, 16 insertions(+), 6 deletions(-) diff --git a/src/playlist/plugins/EmbeddedCuePlaylistPlugin.cxx

Re: [mpd-devel] [PATCH] Workaround: open() ignores 'mode' with O_TMPFILE

2015-05-21 Thread Yclept Nemo
On Thu, May 21, 2015 at 3:13 AM, Max Kellermann m...@duempel.org wrote: What you described are two ways to approach this problem: 1) Always do fchmod() (but with a big fat code comment! And only if building with glibc) 2) do nothing, and if users complain, tell them to upgrade their