RE: [ANNOUNCE] Git v2.17.0

[Randall S. Becker]

On April 2, 2018 3:34 PM, Junio C Hamano wrote:
> Subject: [ANNOUNCE] Git v2.17.0
> 
> The latest feature release Git v2.17.0 is now available at the usual places.  
> It is
> comprised of 516 non-merge commits since v2.16.0, contributed by 71
> people, 20 of which are new faces.

The NonStop platform variant's regression suite (after applying are now very 
small set of platform mods) had the usual failures in t1308:23, t1404:52, and 4 
in t9001. This is equivalent to the same breaks since 2.8.5 through 2.16.2. We 
should be installing in production tomorrow.

Thanks for everyone's hard work.

Cheers,
Randall

-- Brief whoami:
 NonStop developer since approximately 2112884442
 UNIX developer since approximately 421664400
-- In my real life, I talk too much.

RE: [ANNOUNCE] Git v2.17.0

[Randall S. Becker]

On April 2, 2018 7:20 PM, Junio C Hamano wrote:
> To: Stefan Beller <sbel...@google.com>
> Cc: Randall S. Becker <rsbec...@nexbridge.com>; git <git@vger.kernel.org>
> Subject: Re: [ANNOUNCE] Git v2.17.0
> 
> Stefan Beller <sbel...@google.com> writes:
> 
> > Patch at
> > https://public-
> inbox.org/git/010f01d38a9e$a5c4f290$f14ed7b0$@nexbridge
> > .com/
> 
> Thanks for a pointer.  I think it was left behind and got forgotten while
> waiting for a reroll and tying the loose ends.
> 
> I'll go offline for most of the rest of the week.  It would be wonderful
if Git
> 2.17 turns out to be flawless, but that is not a realistic expectation.
Wishing
> for the second best, I'd very much appreciate it if people worked hard to
find
> and fix regressions and collect materials for its first maintenance
release
> 2.17.1 ;-)

No worries. We're running the test suite now (NonStop Platform) and assuming
all is good, it goes into our prod environment this week.

Cheers,
Randall

Re: [ANNOUNCE] Git v2.17.0

[Junio C Hamano]

Stefan Beller  writes:

> Patch at 
> https://public-inbox.org/git/010f01d38a9e$a5c4f290$f14ed7b0$@nexbridge.com/

Thanks for a pointer.  I think it was left behind and got forgotten
while waiting for a reroll and tying the loose ends.

I'll go offline for most of the rest of the week.  It would be
wonderful if Git 2.17 turns out to be flawless, but that is not a
realistic expectation.  Wishing for the second best, I'd very much
appreciate it if people worked hard to find and fix regressions and
collect materials for its first maintenance release 2.17.1 ;-)

Thanks.

RE: [ANNOUNCE] Git v2.17.0

[Randall S. Becker]

On April 2, 2018 4:02 PM, Stefan Beller found my change:
> On Mon, Apr 2, 2018 at 12:57 PM, Randall S. Becker
>  wrote:
> > On April 2, 2018 3:34 PM, Junio C Hamano wrote:
> >> The latest feature release Git v2.17.0 is now available at the usual
> >> places.  It is comprised of 516 non-merge commits since v2.16.0,
> >> contributed by 71 people, 20 of which are new faces.
> >
> > Just a heads up. I think this one might have gotten missed at some point a
> few months back. I think it was submitted back in January. Not sure where it
> fell off or whether it was just dropped.
> >
> > diff --git a/transport-helper.c b/transport-helper.c index
> > 3f380d87d..5ee7007f6 100644
> > --- a/transport-helper.c
> > +++ b/transport-helper.c
> > @@ -1212,7 +1212,7 @@ static int udt_do_read(struct
> unidirectional_transfer *t)
> > return 0;   /* No space for more. */
> >
> > transfer_debug("%s is readable", t->src_name);
> > -   bytes = read(t->src, t->buf + t->bufuse, BUFFERSIZE - t->bufuse);
> > +   bytes = xread(t->src, t->buf + t->bufuse, BUFFERSIZE -
> > + t->bufuse);
> > if (bytes < 0 && errno != EWOULDBLOCK && errno != EAGAIN &&
> > errno != EINTR) {
> > error_errno("read(%s) failed", t->src_name);
> 
> Patch at https://public-
> inbox.org/git/010f01d38a9e$a5c4f290$f14ed7b0$@nexbridge.com/

That was it, thanks.

Cheers,
Randall

Re: [ANNOUNCE] Git v2.17.0

[Stefan Beller]

On Mon, Apr 2, 2018 at 12:57 PM, Randall S. Becker
 wrote:
> On April 2, 2018 3:34 PM, Junio C Hamano wrote:
>> The latest feature release Git v2.17.0 is now available at the usual places. 
>>  It is
>> comprised of 516 non-merge commits since v2.16.0, contributed by 71
>> people, 20 of which are new faces.
>
> Just a heads up. I think this one might have gotten missed at some point a 
> few months back. I think it was submitted back in January. Not sure where it 
> fell off or whether it was just dropped.
>
> diff --git a/transport-helper.c b/transport-helper.c
> index 3f380d87d..5ee7007f6 100644
> --- a/transport-helper.c
> +++ b/transport-helper.c
> @@ -1212,7 +1212,7 @@ static int udt_do_read(struct unidirectional_transfer 
> *t)
> return 0;   /* No space for more. */
>
> transfer_debug("%s is readable", t->src_name);
> -   bytes = read(t->src, t->buf + t->bufuse, BUFFERSIZE - t->bufuse);
> +   bytes = xread(t->src, t->buf + t->bufuse, BUFFERSIZE - t->bufuse);
> if (bytes < 0 && errno != EWOULDBLOCK && errno != EAGAIN &&
> errno != EINTR) {
> error_errno("read(%s) failed", t->src_name);

Patch at 
https://public-inbox.org/git/010f01d38a9e$a5c4f290$f14ed7b0$@nexbridge.com/

RE: [ANNOUNCE] Git v2.17.0

[Randall S. Becker]

On April 2, 2018 3:34 PM, Junio C Hamano wrote:
> The latest feature release Git v2.17.0 is now available at the usual places.  
> It is
> comprised of 516 non-merge commits since v2.16.0, contributed by 71
> people, 20 of which are new faces.

Just a heads up. I think this one might have gotten missed at some point a few 
months back. I think it was submitted back in January. Not sure where it fell 
off or whether it was just dropped.

diff --git a/transport-helper.c b/transport-helper.c
index 3f380d87d..5ee7007f6 100644
--- a/transport-helper.c
+++ b/transport-helper.c
@@ -1212,7 +1212,7 @@ static int udt_do_read(struct unidirectional_transfer *t)
return 0;   /* No space for more. */

transfer_debug("%s is readable", t->src_name);
-   bytes = read(t->src, t->buf + t->bufuse, BUFFERSIZE - t->bufuse);
+   bytes = xread(t->src, t->buf + t->bufuse, BUFFERSIZE - t->bufuse);
if (bytes < 0 && errno != EWOULDBLOCK && errno != EAGAIN &&
errno != EINTR) {
error_errno("read(%s) failed", t->src_name);

Cheers,
Randall

Re: [ANNOUNCE] Git v2.17.0-rc2

[Johannes Schindelin]

Hi team,

On Wed, 28 Mar 2018, Junio C Hamano wrote:

> A release candidate Git v2.17.0-rc2 is now available for testing
> at the usual places.  It is comprised of 499 non-merge commits
> since v2.16.0, contributed by 62 people, 19 of which are new faces.
> 
> I am hoping that we can have the final version tagged at the end of
> coming weekend, before I fly out to Tokyo.  I expect to be offline
> most of the next week after the final is tagged.
> 
> The tarballs are found at:
> 
> https://www.kernel.org/pub/software/scm/git/testing/

... and Git for Windows v2.17.0-rc2 can be found here:

https://github.com/git-for-windows/git/releases/tag/v2.17.0-rc2.windows.1

Ciao,
Johannes

Re: [ANNOUNCE] Git v2.17.0-rc1

[Johannes Schindelin]

Hi Bryan,

On Fri, 23 Mar 2018, Bryan Turner wrote:

> On Fri, Mar 23, 2018 at 10:47 AM, Johannes Schindelin
>  wrote:
> >
> > On Wed, 21 Mar 2018, Junio C Hamano wrote:
> >
> >> A release candidate Git v2.17.0-rc1 is now available for testing
> >> at the usual places.  It is comprised of 493 non-merge commits
> >> since v2.16.0, contributed by 62 people, 19 of which are new faces.
> >>
> >> The tarballs are found at:
> >>
> >> 
> >> https://urldefense.proofpoint.com/v2/url?u=https-3A__www.kernel.org_pub_software_scm_git_testing_=DwIBAg=wBUwXtM9sKhff6UeHOQgvw=uBedA6EFFVX1HiLgmpdrBrv8bIDAScKjk1yk9LOASBM=yXNBIWf9n-gxAIgQyCzXfuKaFkHQaMmwUdtiNBNE8XI=E_Z2M418iwz-HyJg5D0VyTCvyMMd4kGIvYccgJkyTwA=
> >
> > And Git for Windows v2.17.0-rc1 can be found here:
> >
> > https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_git-2Dfor-2Dwindows_git_releases_tag_v2.17.0-2Drc1.windows.1=DwIBAg=wBUwXtM9sKhff6UeHOQgvw=uBedA6EFFVX1HiLgmpdrBrv8bIDAScKjk1yk9LOASBM=yXNBIWf9n-gxAIgQyCzXfuKaFkHQaMmwUdtiNBNE8XI=7ePu15Fwlwuxo8JGcqj-pBNh1wSZYAfYmboqBvJOyA0=
> >
> > Please test so that we can hammer out a robust v2.17.0!
> 
> I've added 2.16.3 and 2.17.0-rc1, for both Linux and Windows, to the
> test matrix for Bitbucket Server. All ~1500 tests have passed for all
> 4 versions.

Thank you so much for testing!

Everybody else: remember that I can only fix bugs pre-emptively in time
for v2.17.0 if you test and report...

Ciao,
Johannes

Re: [ANNOUNCE] Git v2.17.0-rc1

[Derrick Stolee]

On 3/25/2018 2:42 PM, Ævar Arnfjörð Bjarmason wrote:

On Sun, Mar 25 2018, Derrick Stolee wrote:


On 3/23/2018 1:59 PM, Ævar Arnfjörð Bjarmason wrote:

On Wed, Mar 21 2018, Junio C. Hamano wrote:


A release candidate Git v2.17.0-rc1 is now available for testing
at the usual places.  It is comprised of 493 non-merge commits
since v2.16.0, contributed by 62 people, 19 of which are new faces.

I have this deployed on some tens of K machines who all use git in one
way or another (from automated pulls, to users interactively), and rc0
before that, with a few patches on top from me + Takato + Duy + Derrick
since rc0 was released (and since today based on top of rc1). No issues
so far.

The specific in-house version I have is at:
https://github.com/git/git/compare/v2.17.0-rc1...bookingcom:booking-git-v2018-03-23-1

Thanks for testing the commit-graph feature, Ævar! I'm guessing you
have some mechanisms to ensure the 'git commit-graph write' command is
run on these machines and 'core.commitGraph' is set to true in the
config? I would love to hear how this benefits your org.

I haven't deployed any actual use of it at a wider scale, but I've done
some ad-hoc benchmarking with our internal version which has your
patches, and the results are very promising so far on the isolated test
cases where it helps (that you know about, e.g. rev-list --all).

So sorry, I don't have any meaningful testing of this, I just wanted an
easy way to ad-hoc test it & make sure it doesn't break other stuff for
now.

I also threw out most of the manual git maintenance stuff we had and
just rely on gc --auto now, so as soon as you have something to
integrate with that, along with those perf changes Peff suggested I'm
much more likely to play with it in some real way.


Thanks. Integration with 'gc --auto' is a high priority for me after the 
patch lands.


The version on GitHub [1] is slightly ahead of v6 as I wait to reroll on 
v2.17.0. It includes Peff's improvements to inspecting pack-indexes [2].


[1] https://github.com/derrickstolee/git/pull/2
[2] 
https://github.com/derrickstolee/git/pull/2/commits/cb86817ee5c5127b32c93a22ef130f0db6207970

Re: [ANNOUNCE] Git v2.17.0-rc1

[Ævar Arnfjörð Bjarmason]

On Sun, Mar 25 2018, Derrick Stolee wrote:

> On 3/23/2018 1:59 PM, Ævar Arnfjörð Bjarmason wrote:
>> On Wed, Mar 21 2018, Junio C. Hamano wrote:
>>
>>> A release candidate Git v2.17.0-rc1 is now available for testing
>>> at the usual places.  It is comprised of 493 non-merge commits
>>> since v2.16.0, contributed by 62 people, 19 of which are new faces.
>> I have this deployed on some tens of K machines who all use git in one
>> way or another (from automated pulls, to users interactively), and rc0
>> before that, with a few patches on top from me + Takato + Duy + Derrick
>> since rc0 was released (and since today based on top of rc1). No issues
>> so far.
>>
>> The specific in-house version I have is at:
>> https://github.com/git/git/compare/v2.17.0-rc1...bookingcom:booking-git-v2018-03-23-1
>
> Thanks for testing the commit-graph feature, Ævar! I'm guessing you
> have some mechanisms to ensure the 'git commit-graph write' command is
> run on these machines and 'core.commitGraph' is set to true in the
> config? I would love to hear how this benefits your org.

I haven't deployed any actual use of it at a wider scale, but I've done
some ad-hoc benchmarking with our internal version which has your
patches, and the results are very promising so far on the isolated test
cases where it helps (that you know about, e.g. rev-list --all).

So sorry, I don't have any meaningful testing of this, I just wanted an
easy way to ad-hoc test it & make sure it doesn't break other stuff for
now.

I also threw out most of the manual git maintenance stuff we had and
just rely on gc --auto now, so as soon as you have something to
integrate with that, along with those perf changes Peff suggested I'm
much more likely to play with it in some real way.

Re: [ANNOUNCE] Git v2.17.0-rc1

[Derrick Stolee]

On 3/23/2018 1:59 PM, Ævar Arnfjörð Bjarmason wrote:

On Wed, Mar 21 2018, Junio C. Hamano wrote:


A release candidate Git v2.17.0-rc1 is now available for testing
at the usual places.  It is comprised of 493 non-merge commits
since v2.16.0, contributed by 62 people, 19 of which are new faces.

I have this deployed on some tens of K machines who all use git in one
way or another (from automated pulls, to users interactively), and rc0
before that, with a few patches on top from me + Takato + Duy + Derrick
since rc0 was released (and since today based on top of rc1). No issues
so far.

The specific in-house version I have is at:
https://github.com/git/git/compare/v2.17.0-rc1...bookingcom:booking-git-v2018-03-23-1


Thanks for testing the commit-graph feature, Ævar! I'm guessing you have 
some mechanisms to ensure the 'git commit-graph write' command is run on 
these machines and 'core.commitGraph' is set to true in the config? I 
would love to hear how this benefits your org.


Thanks,
-Stolee

Re: [ANNOUNCE] Git v2.17.0-rc1

[Bryan Turner]

On Fri, Mar 23, 2018 at 10:47 AM, Johannes Schindelin
 wrote:
> Hi team,
>
> On Wed, 21 Mar 2018, Junio C Hamano wrote:
>
>> A release candidate Git v2.17.0-rc1 is now available for testing
>> at the usual places.  It is comprised of 493 non-merge commits
>> since v2.16.0, contributed by 62 people, 19 of which are new faces.
>>
>> The tarballs are found at:
>>
>> 
>> https://urldefense.proofpoint.com/v2/url?u=https-3A__www.kernel.org_pub_software_scm_git_testing_=DwIBAg=wBUwXtM9sKhff6UeHOQgvw=uBedA6EFFVX1HiLgmpdrBrv8bIDAScKjk1yk9LOASBM=yXNBIWf9n-gxAIgQyCzXfuKaFkHQaMmwUdtiNBNE8XI=E_Z2M418iwz-HyJg5D0VyTCvyMMd4kGIvYccgJkyTwA=
>
> And Git for Windows v2.17.0-rc1 can be found here:
>
> https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_git-2Dfor-2Dwindows_git_releases_tag_v2.17.0-2Drc1.windows.1=DwIBAg=wBUwXtM9sKhff6UeHOQgvw=uBedA6EFFVX1HiLgmpdrBrv8bIDAScKjk1yk9LOASBM=yXNBIWf9n-gxAIgQyCzXfuKaFkHQaMmwUdtiNBNE8XI=7ePu15Fwlwuxo8JGcqj-pBNh1wSZYAfYmboqBvJOyA0=
>
> Please test so that we can hammer out a robust v2.17.0!

I've added 2.16.3 and 2.17.0-rc1, for both Linux and Windows, to the
test matrix for Bitbucket Server. All ~1500 tests have passed for all
4 versions.

Bryan

Re: [ANNOUNCE] Git v2.17.0-rc1

[Ævar Arnfjörð Bjarmason]

On Wed, Mar 21 2018, Junio C. Hamano wrote:

> A release candidate Git v2.17.0-rc1 is now available for testing
> at the usual places.  It is comprised of 493 non-merge commits
> since v2.16.0, contributed by 62 people, 19 of which are new faces.

I have this deployed on some tens of K machines who all use git in one
way or another (from automated pulls, to users interactively), and rc0
before that, with a few patches on top from me + Takato + Duy + Derrick
since rc0 was released (and since today based on top of rc1). No issues
so far.

The specific in-house version I have is at:
https://github.com/git/git/compare/v2.17.0-rc1...bookingcom:booking-git-v2018-03-23-1

>  * Some bugs around "untracked cache" feature have been fixed.  This
>will notice corrupt data in the untracked cache left by old and
>buggy code and issue a warning---the index can be fixed by clearing
>the untracked cache from it.
>(merge 0cacebf099 nd/fix-untracked-cache-invalidation later to maint).
>(merge 7bf0be7501 ab/untracked-cache-invalidation-docs later to maint).

FYI over some >10k machines I tested on around 1% had (very) verbose
warnings on "status", which was fixed as a one-off with 'git -c
core.untrackedCache=false status' as the
ab/untracked-cache-invalidation-docs suggest, and it hasn't happened
again since.

Re: [ANNOUNCE] Git v2.17.0-rc1

[Johannes Schindelin]

Hi team,

On Wed, 21 Mar 2018, Junio C Hamano wrote:

> A release candidate Git v2.17.0-rc1 is now available for testing
> at the usual places.  It is comprised of 493 non-merge commits
> since v2.16.0, contributed by 62 people, 19 of which are new faces.
> 
> The tarballs are found at:
> 
> https://www.kernel.org/pub/software/scm/git/testing/

And Git for Windows v2.17.0-rc1 can be found here:

https://github.com/git-for-windows/git/releases/tag/v2.17.0-rc1.windows.1

Please test so that we can hammer out a robust v2.17.0!

Ciao,
Johannes

Re: [ANNOUNCE] Git v2.17.0-rc0

[Junio C Hamano]

Junio C Hamano  writes:

> I haven't wordsmithed it fully, but it should say something along
> the lines of ...
>
>  Documentation/RelNotes/2.16.0.txt | 10 ++
>  1 file changed, 10 insertions(+)

Eh, of course the addition should go to 2.17 release notes ;-)  I
just happened to be reviewing a topic forked earlier.

Re: [ANNOUNCE] Git v2.17.0-rc0

[Junio C Hamano]

Ævar Arnfjörð Bjarmason  writes:

> On Fri, Mar 16 2018, Junio C. Hamano jotted:
>
>>   gitweb: hard-depend on the Digest::MD5 5.8 module
>
> I've just noticed this now, but while this module is in 5.8 RedHat's
> butchered perl doesn't have it in the base system, thus this introduces
> the do-we-even-care regression that git's full test suite won't pass on
> a RedHat (or CentOS) base system, because the gitweb tests will fail to
> "use" Digest::MD5.
>
> I'm slightly leaning towards not caring about it, since there's no other
> perl distributor that does this sort of split-out of the core, and if
> you're on a RedHat system they're solving your package problems, so this
> really only impacts the edge case of git developers and redhat
> packagers, both of whom can just do "yum install -y perl-Digest-MD5" to
> fix it.

Thanks for noting.  I agree that this is not something that requires
more than a mention near the beginning of release notes.

I haven't wordsmithed it fully, but it should say something along
the lines of ...

 Documentation/RelNotes/2.16.0.txt | 10 ++
 1 file changed, 10 insertions(+)

diff --git a/Documentation/RelNotes/2.16.0.txt 
b/Documentation/RelNotes/2.16.0.txt
index 8f0461eefd..8b4c24200b 100644
--- a/Documentation/RelNotes/2.16.0.txt
+++ b/Documentation/RelNotes/2.16.0.txt
@@ -6,6 +6,16 @@ Backward compatibility notes and other notable changes.
  * Use of an empty string as a pathspec element that is used for
'everything matches' is now an error.
 
+ * Part of Git that depends on Perl have required at least Perl 5.8
+   since Git v1.7.4 released in 2010, but we used to assume some core
+   modules from Perl distribution may not exist on the system and did
+   a conditional "eval { require <> }"; we no longer do this.
+   On a platform that ships a stripped-down Perl by default, the user
+   may have to install modules the platform chooses not to ship as
+   part of its core (e.g. Digest::MD5, File::Temp, File::Spec,
+   Net::SMTP, NET::Domain).  RedHat/CentOS excludes Digest::MD5 from
+   its base installation, for example.
+
 
 Updates since v2.15
 ---

Re: [ANNOUNCE] Git v2.17.0-rc0

[Ævar Arnfjörð Bjarmason]

On Fri, Mar 16 2018, Junio C. Hamano jotted:

>   gitweb: hard-depend on the Digest::MD5 5.8 module

I've just noticed this now, but while this module is in 5.8 RedHat's
butchered perl doesn't have it in the base system, thus this introduces
the do-we-even-care regression that git's full test suite won't pass on
a RedHat (or CentOS) base system, because the gitweb tests will fail to
"use" Digest::MD5.

I'm slightly leaning towards not caring about it, since there's no other
perl distributor that does this sort of split-out of the core, and if
you're on a RedHat system they're solving your package problems, so this
really only impacts the edge case of git developers and redhat
packagers, both of whom can just do "yum install -y perl-Digest-MD5" to
fix it.