You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Arfrever Frehtes Taifersar Arahesis <ar...@gmail.com> on 2008/09/21 22:06:18 UTC
[RFC] [PATCH] Allow any Neon version from specified branches
I would like to suggest to change Neon version check on Unix-like systems
so that any version from specified branches be allowed.
Currently Subversion downstream maintainers have to update NEON_ALLOWED_LIST
after every new release of Neon (before new version of Subversion is
released which usually contains updated NEON_ALLOWED_LIST).
Neon version check on Windows (build/generator/gen_win.py) is more tolerant.
[[[
Allow any Neon version from specified branches.
* build/ac-macros/neon.m4
(SVN_LIB_NEON, SVN_NEON_CONFIG): Update Neon version checks.
* configure.ac
(NEON_ALLOWED_LIST): Update list.
]]]
--
Arfrever Frehtes Taifersar Arahesis
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Eric Gillespie <ep...@pretzelnet.org>.
Greg Stein <gs...@gmail.com> writes:
> If downstream maintainers are feeling pain, then they can patch out
> our known-good check (not recommended, obviously, but that is their
> call).
It's even easier than that; just use --disable-neon-version-check.
--
Eric Gillespie <*> epg@pretzelnet.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Arfrever Frehtes Taifersar Arahesis <ar...@gmail.com>.
2008-09-22 01:29:18 Greg Stein napisał(a):
> [please review the PATCH mail submission guidelines in hacking.html;
> it was very difficult to review your patch as emailed]
The patch had 'Content-Type: text/x-diff'.
> I don't support this particular patch. It allows 0.27.0 and 0.27.1 to
> be used, which have known regressions. I believe you could alter the
> regex to 0.27.2 in order to fix that.
I fixed it in the new patch.
> Conceptually, I also disagree with the patch, as you, danielsh, and I
> discussed on IRC. In the past, we have had issues with Neon (with all
> dependent libraries, actually). I prefer that we only allow known-good
> libraries, rather than open it to "anything on this branch".
We allow any 1.* version of APR / APR-Util. We shouldn't be specially
restrictive in case of Neon. Neon API/ABI is stable inside specified
branches.
--
Arfrever Frehtes Taifersar Arahesis
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Justin Erenkrantz <ju...@erenkrantz.com>.
On Sun, Sep 21, 2008 at 10:49 PM, Greg Stein <gs...@gmail.com> wrote:
> The mime type meant that it didn't display it... but ugh. text/x-diff
> *is* listed as a valid one. Sigh.
>
> My bad. I thought we said "text/plain" and that was it.
AFAICT, Gmail doesn't let 'text/x-diff' be viewed inline - so in order
to view Arfever's patch, I have to download it - which is a giant
PITA. (So, I haven't reviewed the patch - just followed along on the
reviews so far.)
IMO, we should update hacking to just specify text/plain. Or, figure
out what the right bits are to get the 'view' attachment to work with
Gmail. (Like Outlook, I believe that Gmail is probably used enough by
folks to warrant altering the rules.) -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Greg Stein <gs...@gmail.com>.
The mime type meant that it didn't display it... but ugh. text/x-diff
*is* listed as a valid one. Sigh.
My bad. I thought we said "text/plain" and that was it.
Sorry, Arfrever...
On Sun, Sep 21, 2008 at 10:28 PM, Daniel Shahaf <d....@daniel.shahaf.name> wrote:
> Greg Stein wrote on Sun, 21 Sep 2008 at 21:37 -0700:
>> On Sun, Sep 21, 2008 at 8:33 PM, Peter Samuelson <pe...@p12n.org> wrote:
>> > [Greg Stein]
>> >> [please review the PATCH mail submission guidelines in hacking.html;
>> >> it was very difficult to review your patch as emailed]
>> >
>
> Incredibly non-specific :)
>
> As far as I know, Arfrever did follow the patch submission guidelines.
> Yes, the log message and the patch were separated (one inline and one
> attached), which makes review harder, but that's about the only thing
> I can think of. (The patch even included the 'diff -p' headers, that
> most patches don't have.)
>
>> > Shrug - it came through my MUA inline.
>>
>> Not all MUAs do, which is why we established some guides. When it
>> becomes hard to read, then people simply don't. Arfrever and I had
>> discussed this before, so I was motivated. If it had been for (say)
>> the client lib? Nah. I wouldn't have looked at it, and that is Not
>> Good.
>
> Agreed. But you still don't say *why* you wouldn't have looked at it
> (what made reading it hard).
>
> Daniel
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Greg Stein wrote on Sun, 21 Sep 2008 at 21:37 -0700:
> On Sun, Sep 21, 2008 at 8:33 PM, Peter Samuelson <pe...@p12n.org> wrote:
> > [Greg Stein]
> >> [please review the PATCH mail submission guidelines in hacking.html;
> >> it was very difficult to review your patch as emailed]
> >
Incredibly non-specific :)
As far as I know, Arfrever did follow the patch submission guidelines.
Yes, the log message and the patch were separated (one inline and one
attached), which makes review harder, but that's about the only thing
I can think of. (The patch even included the 'diff -p' headers, that
most patches don't have.)
> > Shrug - it came through my MUA inline.
>
> Not all MUAs do, which is why we established some guides. When it
> becomes hard to read, then people simply don't. Arfrever and I had
> discussed this before, so I was motivated. If it had been for (say)
> the client lib? Nah. I wouldn't have looked at it, and that is Not
> Good.
Agreed. But you still don't say *why* you wouldn't have looked at it
(what made reading it hard).
Daniel
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Greg Stein <gs...@gmail.com>.
On Sun, Sep 21, 2008 at 8:33 PM, Peter Samuelson <pe...@p12n.org> wrote:
> [Greg Stein]
>> [please review the PATCH mail submission guidelines in hacking.html;
>> it was very difficult to review your patch as emailed]
>
> Shrug - it came through my MUA inline.
Not all MUAs do, which is why we established some guides. When it
becomes hard to read, then people simply don't. Arfrever and I had
discussed this before, so I was motivated. If it had been for (say)
the client lib? Nah. I wouldn't have looked at it, and that is Not
Good.
>> In the past, we have had issues with Neon (with all dependent
>> libraries, actually). I prefer that we only allow known-good
>> libraries, rather than open it to "anything on this branch".
>
> We don't do that for other software we use, such as Berkeley DB. And I
> don't think we should. All software has bugs, and I don't think it's
> our responsibility to detect buggy versions of unrelated software.
If we *know* about buggy subsystems, then it is incumbent upon us to
decline to work with those versions.
In general, I believe that I agree with you and Arfrever that maybe we
can be more lax in our version checking on subsystems. With respect to
Neon, we should disable 0.27.{0,1} and not (yet) support 0.29 as you
pointed out.
Cheers,
-g
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Peter Samuelson <pe...@p12n.org>.
[Arfrever Frehtes Taifersar Arahesis]
> What do you think about NEON_ALLOWED_LIST="0\.25 0\.26 0\.27\.2 0\.28 0\.29\.0-dev" ?
I would prefer to see it without those \ in there. I don't think we
need to worry about falsely matching, say, neon "0x26". (Likewise, I'm
not worried about some future neon 0.250.7, which also matches. I hope
neon achieves 1.0.0 before its 250th minor version!)
Also, I am not really in favor of "0.29.0-dev", for the same reason
Greg gives, but that is a separate issue - it's not a regression, since
you already committed that to trunk some time ago.
--
Peter Samuelson | org-tld!p12n!peter | http://p12n.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Greg Stein <gs...@gmail.com>.
A blacklist is not appropriate. We would miss 0.30-dev, etc.
Again. +1 iff references to 0.29 and 0.27.{0,1} are removed.
(and yes, if you push me on inequality, I'm fine with blasting unknown
versions of apr, serf, bdb, sqlite, etc)
On Sep 22, 2008, at 17:19, Daniel Shahaf <d....@daniel.shahaf.name> wrote:
> Arfrever Frehtes Taifersar Arahesis wrote on Tue, 23 Sep 2008 at
> 00:04 +0200:
>> 2008-09-22 23:56:35 Greg Stein napisał(a):
>>> Lose all mention of 0.29, and I'm +1 with that.
>>
>> When I tested Subversion 1.5.2, all tests passed with Neon 0.29.0-
>> dev.
>> http://subversion.tigris.org/servlets/ReadMsg?list=dev&msgNo=142411
>>
>
> Doesn't -dev mean "trunk"? If so, it's pretty much by definition not
> appropriate to list it in NEON_ALLOWED_LIST, which (as I understand
> it)
> means "Versions we tested and we know to work", since it is a moving
> target.
>
> Nonetheless, to support Neon trunk, we can ask those who build with it
> (just Arfrever, right? :)) to pass --disable-neon-version-check, or
> we can
> switch to a blacklist regime (and have NEON_DISALLOWED_LIST="0.27.0 0.27.1
> etc").
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Arfrever Frehtes Taifersar Arahesis wrote on Tue, 23 Sep 2008 at 00:04 +0200:
> 2008-09-22 23:56:35 Greg Stein napisał(a):
> > Lose all mention of 0.29, and I'm +1 with that.
>
> When I tested Subversion 1.5.2, all tests passed with Neon 0.29.0-dev.
> http://subversion.tigris.org/servlets/ReadMsg?list=dev&msgNo=142411
>
Doesn't -dev mean "trunk"? If so, it's pretty much by definition not
appropriate to list it in NEON_ALLOWED_LIST, which (as I understand it)
means "Versions we tested and we know to work", since it is a moving
target.
Nonetheless, to support Neon trunk, we can ask those who build with it
(just Arfrever, right? :)) to pass --disable-neon-version-check, or we can
switch to a blacklist regime (and have NEON_DISALLOWED_LIST="0.27.0 0.27.1
etc").
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Arfrever Frehtes Taifersar Arahesis <ar...@gmail.com>.
2008-09-22 23:56:35 Greg Stein napisał(a):
> Lose all mention of 0.29, and I'm +1 with that.
When I tested Subversion 1.5.2, all tests passed with Neon 0.29.0-dev.
http://subversion.tigris.org/servlets/ReadMsg?list=dev&msgNo=142411
> On Sep 22, 2008, at 15:59, Arfrever Frehtes Taifersar Arahesis <arfrever.fta@gmail.com
> > wrote:
>
> > 2008-09-22 05:33:18 Peter Samuelson napisał(a):
> >> [Greg Stein]
> >>> In the past, we have had issues with Neon (with all dependent
> >>> libraries, actually). I prefer that we only allow known-good
> >>> libraries, rather than open it to "anything on this branch".
> >>
> >> We don't do that for other software we use, such as Berkeley DB.
> >> And I
> >> don't think we should. All software has bugs, and I don't think it's
> >> our responsibility to detect buggy versions of unrelated software.
> >>
> >> I guess the reason we do it for neon is a historical tendency to find
> >> serious bugs and incompatibilities between neon and libsvn_ra_dav.
> >> But
> >> neon has matured a great deal in the past 5+ years. The reason to
> >> whitelist neon versions but not, say, libxml versions is, in my view,
> >> obsolete.
> >>
> >> That said, Arfrever, I don't agree with adding 0.29 to the list of
> >> supported major versions. If it has been tested, I would add it in a
> >> separate commit. 0.29.0-dev is not the same thing, IMO.
> >
> > What do you think about NEON_ALLOWED_LIST="0\.25 0\.26 0\.27\.2 0\.
> > 28 0\.29\.0-dev" ?
--
Arfrever Frehtes Taifersar Arahesis
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Greg Stein <gs...@gmail.com>.
Lose all mention of 0.29, and I'm +1 with that.
On Sep 22, 2008, at 15:59, Arfrever Frehtes Taifersar Arahesis <arfrever.fta@gmail.com
> wrote:
> 2008-09-22 05:33:18 Peter Samuelson napisał(a):
>> [Greg Stein]
>>> In the past, we have had issues with Neon (with all dependent
>>> libraries, actually). I prefer that we only allow known-good
>>> libraries, rather than open it to "anything on this branch".
>>
>> We don't do that for other software we use, such as Berkeley DB.
>> And I
>> don't think we should. All software has bugs, and I don't think it's
>> our responsibility to detect buggy versions of unrelated software.
>>
>> I guess the reason we do it for neon is a historical tendency to find
>> serious bugs and incompatibilities between neon and libsvn_ra_dav.
>> But
>> neon has matured a great deal in the past 5+ years. The reason to
>> whitelist neon versions but not, say, libxml versions is, in my view,
>> obsolete.
>>
>> That said, Arfrever, I don't agree with adding 0.29 to the list of
>> supported major versions. If it has been tested, I would add it in a
>> separate commit. 0.29.0-dev is not the same thing, IMO.
>
> What do you think about NEON_ALLOWED_LIST="0\.25 0\.26 0\.27\.2 0\.
> 28 0\.29\.0-dev" ?
>
> --
> Arfrever Frehtes Taifersar Arahesis
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Arfrever Frehtes Taifersar Arahesis <ar...@gmail.com>.
2008-09-22 05:33:18 Peter Samuelson napisał(a):
> [Greg Stein]
> > In the past, we have had issues with Neon (with all dependent
> > libraries, actually). I prefer that we only allow known-good
> > libraries, rather than open it to "anything on this branch".
>
> We don't do that for other software we use, such as Berkeley DB. And I
> don't think we should. All software has bugs, and I don't think it's
> our responsibility to detect buggy versions of unrelated software.
>
> I guess the reason we do it for neon is a historical tendency to find
> serious bugs and incompatibilities between neon and libsvn_ra_dav. But
> neon has matured a great deal in the past 5+ years. The reason to
> whitelist neon versions but not, say, libxml versions is, in my view,
> obsolete.
>
> That said, Arfrever, I don't agree with adding 0.29 to the list of
> supported major versions. If it has been tested, I would add it in a
> separate commit. 0.29.0-dev is not the same thing, IMO.
What do you think about NEON_ALLOWED_LIST="0\.25 0\.26 0\.27\.2 0\.28 0\.29\.0-dev" ?
--
Arfrever Frehtes Taifersar Arahesis
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Peter Samuelson <pe...@p12n.org>.
[Greg Stein]
> [please review the PATCH mail submission guidelines in hacking.html;
> it was very difficult to review your patch as emailed]
Shrug - it came through my MUA inline.
> In the past, we have had issues with Neon (with all dependent
> libraries, actually). I prefer that we only allow known-good
> libraries, rather than open it to "anything on this branch".
We don't do that for other software we use, such as Berkeley DB. And I
don't think we should. All software has bugs, and I don't think it's
our responsibility to detect buggy versions of unrelated software.
I guess the reason we do it for neon is a historical tendency to find
serious bugs and incompatibilities between neon and libsvn_ra_dav. But
neon has matured a great deal in the past 5+ years. The reason to
whitelist neon versions but not, say, libxml versions is, in my view,
obsolete.
That said, Arfrever, I don't agree with adding 0.29 to the list of
supported major versions. If it has been tested, I would add it in a
separate commit. 0.29.0-dev is not the same thing, IMO.
--
Peter Samuelson | org-tld!p12n!peter | http://p12n.org/
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Greg Stein <gs...@gmail.com>.
[please review the PATCH mail submission guidelines in hacking.html;
it was very difficult to review your patch as emailed]
I don't support this particular patch. It allows 0.27.0 and 0.27.1 to
be used, which have known regressions. I believe you could alter the
regex to 0.27.2 in order to fix that.
Conceptually, I also disagree with the patch, as you, danielsh, and I
discussed on IRC. In the past, we have had issues with Neon (with all
dependent libraries, actually). I prefer that we only allow known-good
libraries, rather than open it to "anything on this branch".
If downstream maintainers are feeling pain, then they can patch out
our known-good check (not recommended, obviously, but that is their
call).
Cheers,
-g
On Sun, Sep 21, 2008 at 3:06 PM, Arfrever Frehtes Taifersar Arahesis
<ar...@gmail.com> wrote:
> I would like to suggest to change Neon version check on Unix-like systems
> so that any version from specified branches be allowed.
>
> Currently Subversion downstream maintainers have to update NEON_ALLOWED_LIST
> after every new release of Neon (before new version of Subversion is
> released which usually contains updated NEON_ALLOWED_LIST).
>
> Neon version check on Windows (build/generator/gen_win.py) is more tolerant.
>
> [[[
> Allow any Neon version from specified branches.
>
> * build/ac-macros/neon.m4
> (SVN_LIB_NEON, SVN_NEON_CONFIG): Update Neon version checks.
>
> * configure.ac
> (NEON_ALLOWED_LIST): Update list.
> ]]]
>
> --
> Arfrever Frehtes Taifersar Arahesis
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: [RFC] [PATCH] Allow any Neon version from specified branches
Posted by Arfrever Frehtes Taifersar Arahesis <ar...@gmail.com>.
2008-09-22 00:06:18 Arfrever Frehtes Taifersar Arahesis napisał(a):
> I would like to suggest to change Neon version check on Unix-like systems
> so that any version from specified branches be allowed.
>
> Currently Subversion downstream maintainers have to update NEON_ALLOWED_LIST
> after every new release of Neon (before new version of Subversion is
> released which usually contains updated NEON_ALLOWED_LIST).
>
> Neon version check on Windows (build/generator/gen_win.py) is more tolerant.
>
> [[[
> Allow any Neon version from specified branches.
>
> * build/ac-macros/neon.m4
> (SVN_LIB_NEON, SVN_NEON_CONFIG): Update Neon version checks.
>
> * configure.ac
> (NEON_ALLOWED_LIST): Update list.
> ]]]
>
Committed in r33272.
--
Arfrever Frehtes Taifersar Arahesis