rc1 planning

rc1 planning

[Karel Zak]

 Hi,

 I'd like to release v2.28-rc1 ASAP, but the rest of this week and all
 next week I have vacation (and all next week I'm going to be off-line
 somewhere in Atlantic). So, my plan is 7th or 8th March.

 remaining issues:

  * btrfs loop devs in libmount

  * libblkid vs. CDROMs (check if last two 2K sectors are readable, if
    not reduce probing area)    [me]

  * add time formatting functions for ISO and ctime [not urgent]
 
  * cleanup openat() related function [not urgent, Rudi?]

  * add --zeroout to blkdiscard [me]

  * fix look(1) [https://github.com/karelzak/util-linux/issues/284]

  * cleanup docs, gtk-doc warnings, etc.

  * ???


    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

Re: rc1 planning

[Ruediger Meier]

On Wednesday 24 February 2016, Karel Zak wrote:
>  Hi,
>
>  I'd like to release v2.28-rc1 ASAP, but the rest of this week and
> all next week I have vacation (and all next week I'm going to be
> off-line somewhere in Atlantic). So, my plan is 7th or 8th March.
>
>  remaining issues:
>
>   * btrfs loop devs in libmount
>
>   * libblkid vs. CDROMs (check if last two 2K sectors are readable,
> if not reduce probing area)    [me]
>
>   * add time formatting functions for ISO and ctime [not urgent]
>
>   * cleanup openat() related function [not urgent, Rudi?]

I've changed my mind a bit about lib/at.c removal because it's used by 
sysfs.c and we would lose libblkid and fdisk on non-linux systems. I 
will review first whether we could build libblkid without sysfs for 
non-Linux.

Regarding my stupid commit ...
  8b9cf26a "lib: fix unused parameters and variables"
... next time I would try to add our new "NO_UNUSED_WARN_CFLAGS" just 
for single source files. Not using __unused macros would keep our 
heavily ifdef'ed sources more readable.

>   * add --zeroout to blkdiscard [me]
>
>   * fix look(1) [https://github.com/karelzak/util-linux/issues/284]
>
>   * cleanup docs, gtk-doc warnings, etc.

My favorite warning one could fix is this one:

Making all in libblkid/docs
make[2]: Entering directory `/home/rudi/devel/util-linux/libblkid/docs'
make[2]: Circular libblkid-sections.txt <- scan-build.stamp dependency 
dropped.

>   * ???
>
>
>     Karel

Re: rc1 planning

[Karel Zak]

On Wed, Feb 24, 2016 at 03:50:21PM +0100, Ruediger Meier wrote:
> On Wednesday 24 February 2016, Karel Zak wrote:
> >  Hi,
> >
> >  I'd like to release v2.28-rc1 ASAP, but the rest of this week and
> > all next week I have vacation (and all next week I'm going to be
> > off-line somewhere in Atlantic). So, my plan is 7th or 8th March.
> >
> >  remaining issues:
> >
> >   * btrfs loop devs in libmount
> >
> >   * libblkid vs. CDROMs (check if last two 2K sectors are readable,
> > if not reduce probing area)    [me]
> >
> >   * add time formatting functions for ISO and ctime [not urgent]
> >
> >   * cleanup openat() related function [not urgent, Rudi?]
> 
> I've changed my mind a bit about lib/at.c removal because it's used by 
> sysfs.c and we would lose libblkid and fdisk on non-linux systems. I 
> will review first whether we could build libblkid without sysfs for 
> non-Linux.

 The libblkid is optional for libfdisk, I cannot imagine libblkid without
 sysfs :-)

> Regarding my stupid commit ...
>   8b9cf26a "lib: fix unused parameters and variables"
> ... next time I would try to add our new "NO_UNUSED_WARN_CFLAGS" just 
> for single source files. Not using __unused macros would keep our 
> heavily ifdef'ed sources more readable.
> 
> >   * add --zeroout to blkdiscard [me]
> >
> >   * fix look(1) [https://github.com/karelzak/util-linux/issues/284]
> >
> >   * cleanup docs, gtk-doc warnings, etc.
> 
> My favorite warning one could fix is this one:

Frankly, my favorite solution is to drop gtk-docs and use something
else. I like gtk-docs html output, but the rest (warnings, docs
formatting, Makefile integration, ...) is painful.

    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

[PATCH] setarch: fix personality syscall return code check

[Dmitry V. Levin]

Depending on architecture and kernel version, personality
syscall is either capable or incapable of returning an error.
If the return value is not an error, then it's the previous
personality value, which can be an arbitrary value
undistinguishable from an error value.
To make things clear, a second call is needed.

For more details about personality syscall peculiarities see
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=glibc-2.22-637-ge0043e17dfc5

Signed-off-by: Dmitry V. Levin <ldv@altlinux.org>
---
 sys-utils/setarch.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/sys-utils/setarch.c b/sys-utils/setarch.c
index b6159c5..b5e401c 100644
--- a/sys-utils/setarch.c
+++ b/sys-utils/setarch.c
@@ -236,8 +236,18 @@ static int set_arch(const char *pers, unsigned long options, int list)
 	if (transitions[i].perval < 0)
 		errx(EXIT_FAILURE, _("%s: Unrecognized architecture"), pers);
 	pers_value = transitions[i].perval | options;
-	if (personality(pers_value) == -EINVAL)
-		return 1;
+	if (personality(pers_value) < 0) {
+		/*
+		 * Depending on architecture and kernel version, personality
+		 * syscall is either capable or incapable of returning an error.
+		 * If the return value is not an error, then it's the previous
+		 * personality value, which can be an arbitrary value
+		 * undistinguishable from an error value.
+		 * To make things clear, a second call is needed.
+		 */
+		if (personality(pers_value) < 0)
+			return 1;
+	}
 	uname(&un);
 	if (transitions[i].result_arch && strcmp(un.machine, transitions[i].result_arch)) {
 		if (strcmp(transitions[i].result_arch, "i386")

-- 
ldv

Re: [PATCH] setarch: fix personality syscall return code check

[Karel Zak]

On Sat, Mar 05, 2016 at 12:22:52AM +0300, Dmitry V. Levin wrote:
>  sys-utils/setarch.c | 14 ++++++++++++--
>  1 file changed, 12 insertions(+), 2 deletions(-)

Nice. Applied, thanks.

    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com