[blfs-dev] glibc errors in blfs packages on i686

Bruce Dubbs bruce.dubbs at gmail.com
Thu Aug 29 14:37:53 PDT 2013


Ken Moffat wrote:
> On Thu, Aug 29, 2013 at 02:43:30PM -0500, Bruce Dubbs wrote:
>> Ken Moffat wrote:
>>> On Thu, Aug 29, 2013 at 10:02:41AM -0500, Bruce Dubbs wrote:
>>>>
>>>> OK, I'll do a -rc2 today.  I think it best to slip the stable release to
>>>> September 7.  Should we update to the newer gettext, kbd, and/or kmod at
>>>> the same time?  Those all seem to have lower potential for problems than
>>>> the glibc sed.
>>
>>>    I'm not in favour of updating anything except glibc.
>>
>> I understand your point of view.  Looking at things from a slightly
>> different perspective, I don't use kbd at all.  kmod is pretty much a
>> non-issue for me because I really minimize modules.  Right now on my
>> development system, none are loaded.  It will load SCSI_DEBUG when
>> running util-linux tests.  Looking at gettext, the release in the book
>> is from June.  From the latest release changelog:
>>
>>    * windows/gettext.rc: Update copyright year.
>>    * NEWS: Mention libasprintf change and Tcl bug fix.
>>
>> That's it.
>>
>> The idea of including these is that there is a low probability of the
>> newer packages causing problems and even if there are, the impact would
>> be small.
>>
>> I will update to the latest kernel.  We already say in the book to do
>> that.
>>
>> Those are my observations.  If we need to retest, testing everything at
>> once would be easier.
>>
>> All that said, I'm willing to do what we collectively think best.
>>
>>     -- Bruce
>
>   Sorry it's taken me a while to reply, I've been doing a different
> sort of tagging with metaflac.
>
>   For kbd, I guess the most likely problem would be a
> less-than-optimal set of instructions.  As I noted in the ticket,
> I'm undecided whether a separate libkbd.so is worthwhile or not
> worthwhile.
>
>   For kmod, I guess it works or it doesn't.  NB we warn about running
> the util-linux tests as root, but running as a user doesn't load the
> SCSI_DEBUG module (I tested as a user on my i686, without enabling
> that kernel option).
>
>   Certainly, the gettext changes do sound harmless.
>
>   OTOH the more we change for newer and shinier versions, the less
> reason people have for respecting the -rc name.

But we are discussing -rc2, not a stable release.  If we didn't have the 
glibc issue, I wouldn't be suggesting the other packages.

>   Are we going to keep the 7.4 tags showing 7.4 ?  On the grounds
> that whatever changes in -rc2 is _unlikely_ to break BLFS, I suppose
> we could leave them.  Unless someone finds a problem.

I think the issue was that we were doing the tests for -rc1 and found 
the glibc error.  -rc2 would continue to do the testing.  We didn't find 
the issue until Fernando tried java on a i686.  I could do a global sed 
and replace lfs74 with lfs74rc1 and then start all of BLFS over.

Right now I'm thinking about a potential BLFS release in mid-October or 
early November.  After 5 years, what's a month or two?

   -- Bruce






More information about the blfs-dev mailing list