You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Otto Kolsi <ot...@kolsi.fi> on 2008/09/18 11:44:20 UTC
SVN 1.5.2 and Solaris 10 make check error with prop_tests.py 22
Hi all,
With help of "danielsh" in IRC, we found following. In Solaris 10
(SPARC) svn 1.5.1 "make check" goes well. With 1.5.2 one test fails,
this can be verified by running "./prop_tests.py 22". Failing test can
be seen here: http://paste.lisp.org/display/67064
It seems that between 1.5.1 and 1.5.2 there was this modification:
http://svn.collab.net/viewvc-experimental/svn/?view=revision&revision=32491
Failing test can be fixed by commenting the LC_ALL line in svntest/main.py.
I'm holding back the upgrade from 1.5.1 to 1.5.2 in our production
environment because of this. Is this failing test significant and is it
safe to upgrade? Can this be resolved somehow in future versions?
P.S. I'm not subscribed to the dev-list
--
Otto
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: SVN 1.5.2 and Solaris 10 make check error with prop_tests.py
22
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Jens Seidel wrote on Fri, 19 Sep 2008 at 18:24 +0200:
> I always wonder about the info documentation in libc:
>
> `LC_ALL'
> This is not an environment variable; it is only a macro that you
> can use with `setlocale' to set a single locale for all purposes.
> Setting this environment variable overwrites all selections by the
> other `LC_*' variables or `LANG'.
>
> The first part says it is not an environment variable, the second part
> says the opposite.
>
According to svn_cmdline_init(), it is both a macro and an env var.
Daniel
> a) Otto, what happens if you change the test to use LANG or LANGUAGE instead
> of LC_ALL?
> b) Does something change if you use fi_FI.ISO8859-1 for all LC_*
> entries (try LANG=fi_FI.ISO8859-1)?
>
> Jens
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: SVN 1.5.2 and Solaris 10 make check error with prop_tests.py
22
Posted by Otto Kolsi <ot...@kolsi.fi>.
Sorry for the delay, I can do tests related to this issue only during
"working hours".
Jens Seidel wrote:
> On Fri, Sep 19, 2008 at 02:42:51PM +0300, Otto Kolsi wrote:
>> Julian Foad wrote:
>>>> If you would like to help trace the cause of the problem, that would be
>>>> great. (What local do you use normally?)
>> I can try to help as much as I can. When executing "locale" in the
>> Solaris machine I get:
>>
>> LANG=
>> LC_CTYPE=fi_FI.ISO8859-1
>> LC_NUMERIC=fi_FI.ISO8859-1
>> LC_TIME="C"
>> LC_COLLATE=fi_FI.ISO8859-1
>> LC_MONETARY=fi_FI.ISO8859-1
>> LC_MESSAGES=C
>> LC_ALL=
>>
>> I'm not that familiar with locales.. not sure if it is okay or not to
>> have e.g. empty LC_ALL locale.
>
> Please confirm that the variable LANGUAGE is not set.
I can confirm that LANGUAGE is not set.
> I always wonder about the info documentation in libc:
>
> `LC_ALL'
> This is not an environment variable; it is only a macro that you
> can use with `setlocale' to set a single locale for all purposes.
> Setting this environment variable overwrites all selections by the
> other `LC_*' variables or `LANG'.
>
> The first part says it is not an environment variable, the second part
> says the opposite.
>
> a) Otto, what happens if you change the test to use LANG or LANGUAGE instead
> of LC_ALL?
If I set LANG = 'C' or LANGUAGE = 'C' in the main.py, test works.
> b) Does something change if you use fi_FI.ISO8859-1 for all LC_*
> entries (try LANG=fi_FI.ISO8859-1)?
If I do "export LANG=fi_FI.ISO8859-1" in shell and confirm with "locale"
that it is set, the test does not work. Also if I similarly set LC_ALL
to the FI code, test does not work.
And also, setting either LANG or LANGUAGE in shell does not make the
test work (at the same time the main.py sets LC_ALL to C).
--
Otto
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: SVN 1.5.2 and Solaris 10 make check error with prop_tests.py 22
Posted by Jens Seidel <je...@users.sourceforge.net>.
On Fri, Sep 19, 2008 at 02:42:51PM +0300, Otto Kolsi wrote:
> Julian Foad wrote:
> >>If you would like to help trace the cause of the problem, that would be
> >>great. (What local do you use normally?)
>
> I can try to help as much as I can. When executing "locale" in the
> Solaris machine I get:
>
> LANG=
> LC_CTYPE=fi_FI.ISO8859-1
> LC_NUMERIC=fi_FI.ISO8859-1
> LC_TIME="C"
> LC_COLLATE=fi_FI.ISO8859-1
> LC_MONETARY=fi_FI.ISO8859-1
> LC_MESSAGES=C
> LC_ALL=
>
> I'm not that familiar with locales.. not sure if it is okay or not to
> have e.g. empty LC_ALL locale.
Please confirm that the variable LANGUAGE is not set.
I always wonder about the info documentation in libc:
`LC_ALL'
This is not an environment variable; it is only a macro that you
can use with `setlocale' to set a single locale for all purposes.
Setting this environment variable overwrites all selections by the
other `LC_*' variables or `LANG'.
The first part says it is not an environment variable, the second part
says the opposite.
a) Otto, what happens if you change the test to use LANG or LANGUAGE instead
of LC_ALL?
b) Does something change if you use fi_FI.ISO8859-1 for all LC_*
entries (try LANG=fi_FI.ISO8859-1)?
Jens
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: SVN 1.5.2 and Solaris 10 make check error with prop_tests.py
22
Posted by Otto Kolsi <ot...@kolsi.fi>.
Julian Foad wrote:
>> 1. To ensure you can delete a property whose name contains invalid
>> characters.
>>
>> 2. To ensure you cannot create such a property.
>>
>> Part (1) is failing. "svn" refuses to delete a property whose name
>> contains an invalid character.
>>
>> I don't know exactly why it fails for you in the "C" locale and not in
>> another locale, or what the correct fix would be, but I can say that the
>> only reason for wanting behaviour (1) is so that you can clean up your
>> properties if you somehow got a bad property name (maybe when using a
>> client that didn't check for a valid name).
>>
>> This is not an essential part of Subversion's operation, so you can
>> safely ignore that failure.
Thank you for the information, I can now do the upgrade to 1.5.2 :)
>> If you would like to help trace the cause of the problem, that would be
>> great. (What local do you use normally?)
I can try to help as much as I can. When executing "locale" in the
Solaris machine I get:
LANG=
LC_CTYPE=fi_FI.ISO8859-1
LC_NUMERIC=fi_FI.ISO8859-1
LC_TIME="C"
LC_COLLATE=fi_FI.ISO8859-1
LC_MONETARY=fi_FI.ISO8859-1
LC_MESSAGES=C
LC_ALL=
I'm not that familiar with locales.. not sure if it is okay or not to
have e.g. empty LC_ALL locale.
--
Otto
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: SVN 1.5.2 and Solaris 10 make check error with prop_tests.py 22
Posted by Julian Foad <ju...@btopenworld.com>.
On Fri, 2008-09-19 at 09:52 +0100, Julian Foad wrote:
> On Thu, 2008-09-18 at 14:44 +0300, Otto Kolsi wrote:
> > With help of "danielsh" in IRC, we found following. In Solaris 10
> > (SPARC) svn 1.5.1 "make check" goes well. With 1.5.2 one test fails,
> > this can be verified by running "./prop_tests.py 22". Failing test can
> > be seen here: http://paste.lisp.org/display/67064
>
> Otto,
>
> I examined the failure. The purpose of that test is two-fold:
BTW, I mean I examined the output you pasted, and the code that
generates it. I was not able to reproduce this failure myself.
- Julian
> 1. To ensure you can delete a property whose name contains invalid
> characters.
>
> 2. To ensure you cannot create such a property.
>
> Part (1) is failing. "svn" refuses to delete a property whose name
> contains an invalid character.
>
> I don't know exactly why it fails for you in the "C" locale and not in
> another locale, or what the correct fix would be, but I can say that the
> only reason for wanting behaviour (1) is so that you can clean up your
> properties if you somehow got a bad property name (maybe when using a
> client that didn't check for a valid name).
>
> This is not an essential part of Subversion's operation, so you can
> safely ignore that failure.
>
> If you would like to help trace the cause of the problem, that would be
> great. (What local do you use normally?)
>
> Thanks.
> - Julian
>
>
> > It seems that between 1.5.1 and 1.5.2 there was this modification:
> > http://svn.collab.net/viewvc-experimental/svn/?view=revision&revision=32491
> >
> > Failing test can be fixed by commenting the LC_ALL line in svntest/main.py.
> >
> > I'm holding back the upgrade from 1.5.1 to 1.5.2 in our production
> > environment because of this. Is this failing test significant and is it
> > safe to upgrade? Can this be resolved somehow in future versions?
> >
> > P.S. I'm not subscribed to the dev-list
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: dev-help@subversion.tigris.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: SVN 1.5.2 and Solaris 10 make check error with prop_tests.py 22
Posted by Julian Foad <ju...@btopenworld.com>.
On Thu, 2008-09-18 at 14:44 +0300, Otto Kolsi wrote:
> With help of "danielsh" in IRC, we found following. In Solaris 10
> (SPARC) svn 1.5.1 "make check" goes well. With 1.5.2 one test fails,
> this can be verified by running "./prop_tests.py 22". Failing test can
> be seen here: http://paste.lisp.org/display/67064
Otto,
I examined the failure. The purpose of that test is two-fold:
1. To ensure you can delete a property whose name contains invalid
characters.
2. To ensure you cannot create such a property.
Part (1) is failing. "svn" refuses to delete a property whose name
contains an invalid character.
I don't know exactly why it fails for you in the "C" locale and not in
another locale, or what the correct fix would be, but I can say that the
only reason for wanting behaviour (1) is so that you can clean up your
properties if you somehow got a bad property name (maybe when using a
client that didn't check for a valid name).
This is not an essential part of Subversion's operation, so you can
safely ignore that failure.
If you would like to help trace the cause of the problem, that would be
great. (What local do you use normally?)
Thanks.
- Julian
> It seems that between 1.5.1 and 1.5.2 there was this modification:
> http://svn.collab.net/viewvc-experimental/svn/?view=revision&revision=32491
>
> Failing test can be fixed by commenting the LC_ALL line in svntest/main.py.
>
> I'm holding back the upgrade from 1.5.1 to 1.5.2 in our production
> environment because of this. Is this failing test significant and is it
> safe to upgrade? Can this be resolved somehow in future versions?
>
> P.S. I'm not subscribed to the dev-list
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org