Re: New branch and feature freeze for Lucene/Solr 8.10.0
: I was referring to doing this with languages other than java.
:
: I'm also assuming that exceeding this limit is going to cause indirect
: hassles for users of lucene, e.g. breaking various security / supply
: chain tools. We know a lot of these are total crap but people in the
: corporate world have to suffer under them.
Just to be clear -- our 'Implementation-Version:' has been exceeding the
72 byte "single line" limit for a LONG time -- worrying about how
"security / supply chain" tools will handle parsing that line now seems
kind of silly...
If tools can't handle a line wrap in the 8.10 jars, then they haven't
been able to handle the line wrap since we switched from svn to git (when
the Implementation Version values switched from being based svn version#
to git sha)
The *ONLY* thing that's new here is where in the value the line wrap
happens (with 8.10.0 it happens in the middle of the SHA) and that our
smoketest tool isn't smart enough to parse the values properly.
This is not even the first time we've even had a conversation about the
smoke tester and Implementation Version line wraps: LUCENE-7023.
: Its super-easy to use a short hash here and avoid problems.
There is however an actual and practical downside to switching our
implementation version to using a "short" SHA, and that's that we would
lose the ability to garuntee that the information in the
Implementation-Version uniquely identifies what commit a given jar was
built from. Multiple commits with the same short(end) hash are possible
-- Multiple commits with identical (full) commits is not.
Folks may think that using git tags is useful enough for figuring this
out from official releases, but being able to look at the jar metadata
from arbitrary builds off of arbitrary branches and sanity check where
exactly they come from has been very useful to me for 10+ years.
: On Thu, Sep 16, 2021 at 3:03 AM Dawid Weiss wrote:
: >
: > Jar command doesn't have it, true. But it's fairly trivial to do, even
: > with an inline snippet like this one?
: >
: > public class PrintManifest {
: > public static void main(String[] jars) throws IOException {
: > for (var jar : jars) {
: > var manifest = new JarFile(Paths.get(jar).toFile()).getManifest();
: > var attrs = manifest.getMainAttributes();
: > System.out.println(jar + ": " +
attrs.getValue("Implementation-Version"));
: > }
: > }
: > }
: >
: > I have this in my lucene-core-9.0.0-SNAPSHOT.jar:
: >
: > Implementation-Version: 9.0.0-SNAPSHOT de45b68c909815ce5ea7b6b9e1a2ce337
: > 5b6334d [snapshot build, details omitted]
: >
: > and running:
: >
: > java PrintManifest.java lucene-core-9.0.0-SNAPSHOT.jar
: >
: > shows:
: >
: > lucene-core-9.0.0-SNAPSHOT.jar: 9.0.0-SNAPSHOT
: > de45b68c909815ce5ea7b6b9e1a2ce3375b6334d [snapshot build, details
: > omitted]
: >
: > This seems easier to me than trying to remember and keep the length of
: > that line shorter than an arbitrary limit.
: >
: > Dawid
: >
: >
: > On Wed, Sep 15, 2021 at 9:46 PM Robert Muir wrote:
: > >
: > > But its irrelevant that is "valid" when virtually no tools match it.
: > >
: > > In other words, I'd agree with you if the "jar" command had some
: > > ability to read these manifests and print stuff to stdout, e.g. if
: > > there was ANY interop at all here.
: > >
: > > But there isn't. So IMO it makes no sense to cause confusion and chaos
: > > by adding an unnecessarily long git commit hash.
: > >
: > > On Wed, Sep 15, 2021 at 3:26 PM Dawid Weiss wrote:
: > > >
: > > >
: > > > This is valid manifest line-breaking though... Can we read the manifest
properly on the smoke tester side somehow (for example, run a Java process that
reads and extracts the required attribute)? This way we wouldn't care about the
implementation details of how manifest wraps the lines (or escapes characters).
: > > >
: > > > D.
: > > >
: > > > On Wed, Sep 15, 2021 at 8:46 PM Mike Drob wrote:
: > > >>
: > > >> The benchmark jar has the info we need… sort of. When I built it, it
has:
: > > >>
: > > >> Implementation-Version: 8.10.0 75a5061d3715cc5d93c4cbe4f1fa62bf035eea1
: > > >> 1 - mdrob - 2021-09-15 11:40:36
: > > >>
: > > >>
: > > >> and it’s looking for Implementation-Version: 8.10.0
75a5061d3715cc5d93c4cbe4f1fa62bf035eea11 on one line.
: > > >>
: > > >> Because 8.10 is a character longer than 8.9, we happen to wrap the
last character of the git commit sha. From the manifest spec:
: > > >>
: > > >> No line may be longer than 72 bytes (not characters), in its
UTF8-encoded form.
: > > >> If a value would make the initial line longer than this, it should be
continued
: > > >> on extra lines (each starting with a single SPACE).
: > > >>
: > > >> And we were already teetering on the edge of that limit. We'll run
into this problem again in a few years when we try to release version 10.0.0,
so solving it now has practical benefits down the line.
: > > >>
: > > >> There's a few options that I can
[jira] [Closed] (PYLUCENE-60) we are providing the projects for mtech and btech students.
[ https://issues.apache.org/jira/browse/PYLUCENE-60?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Andi Vajda closed PYLUCENE-60. -- Resolution: Invalid This is spam, right ? > we are providing the projects for mtech and btech students. > --- > > Key: PYLUCENE-60 > URL: https://issues.apache.org/jira/browse/PYLUCENE-60 > Project: PyLucene > Issue Type: Blog - New Blog Request > Environment: python >Reporter: Nasima Takeoff >Priority: Trivial > Labels: btech, final, mtech, projects, python > Attachments: Python projects.png > > > h4. Python has been within the top 10 popular programming languages for an > extended time because the community of Python programmers has grown tons > thanks to its easy syntax and library support. during this article, Takeoff > will be able to introduce you to 60 amazing Python projects with ASCII text > files solved and explained by experts in [Takeoff > Projects|https://takeoffprojects.com/]. -- This message was sent by Atlassian Jira (v8.3.4#803005)
Re: New branch and feature freeze for Lucene/Solr 8.10.0
Ok, fine.
D.
On Thu, Sep 16, 2021 at 12:00 PM Robert Muir wrote:
>
> The shit that mike's python tool is doing, i'd expect bugs like that.
> If you honestly disagree with me, go fix his python code to obey "the
> rules". Be sure you have "the rules" correct as well (count bytes, not
> characters, count EOL, etc).
>
> Try to imagine how much software is going to get this wrong. Various
> tools scanning for security problems that aren't written in java and
> hence aren't using Jar class. Your security team isn't writing fucking
> java code. Your ops team isn't writing fucking java code, so it really
> does not matter if it is "easy in java".
>
> This is *exactly* what i mean when i say this crazy shit has NO
> INTEROP. It is 100% pure garbage, java showing its true colors.
>
> On Thu, Sep 16, 2021 at 5:53 AM Robert Muir wrote:
> >
> > I was referring to doing this with languages other than java.
> >
> > I'm also assuming that exceeding this limit is going to cause indirect
> > hassles for users of lucene, e.g. breaking various security / supply
> > chain tools. We know a lot of these are total crap but people in the
> > corporate world have to suffer under them.
> >
> > Its super-easy to use a short hash here and avoid problems.
> >
> > On Thu, Sep 16, 2021 at 3:03 AM Dawid Weiss wrote:
> > >
> > > Jar command doesn't have it, true. But it's fairly trivial to do, even
> > > with an inline snippet like this one?
> > >
> > > public class PrintManifest {
> > > public static void main(String[] jars) throws IOException {
> > > for (var jar : jars) {
> > > var manifest = new JarFile(Paths.get(jar).toFile()).getManifest();
> > > var attrs = manifest.getMainAttributes();
> > > System.out.println(jar + ": " +
> > > attrs.getValue("Implementation-Version"));
> > > }
> > > }
> > > }
> > >
> > > I have this in my lucene-core-9.0.0-SNAPSHOT.jar:
> > >
> > > Implementation-Version: 9.0.0-SNAPSHOT de45b68c909815ce5ea7b6b9e1a2ce337
> > > 5b6334d [snapshot build, details omitted]
> > >
> > > and running:
> > >
> > > java PrintManifest.java lucene-core-9.0.0-SNAPSHOT.jar
> > >
> > > shows:
> > >
> > > lucene-core-9.0.0-SNAPSHOT.jar: 9.0.0-SNAPSHOT
> > > de45b68c909815ce5ea7b6b9e1a2ce3375b6334d [snapshot build, details
> > > omitted]
> > >
> > > This seems easier to me than trying to remember and keep the length of
> > > that line shorter than an arbitrary limit.
> > >
> > > Dawid
> > >
> > >
> > > On Wed, Sep 15, 2021 at 9:46 PM Robert Muir wrote:
> > > >
> > > > But its irrelevant that is "valid" when virtually no tools match it.
> > > >
> > > > In other words, I'd agree with you if the "jar" command had some
> > > > ability to read these manifests and print stuff to stdout, e.g. if
> > > > there was ANY interop at all here.
> > > >
> > > > But there isn't. So IMO it makes no sense to cause confusion and chaos
> > > > by adding an unnecessarily long git commit hash.
> > > >
> > > > On Wed, Sep 15, 2021 at 3:26 PM Dawid Weiss
> > > > wrote:
> > > > >
> > > > >
> > > > > This is valid manifest line-breaking though... Can we read the
> > > > > manifest properly on the smoke tester side somehow (for example, run
> > > > > a Java process that reads and extracts the required attribute)? This
> > > > > way we wouldn't care about the implementation details of how manifest
> > > > > wraps the lines (or escapes characters).
> > > > >
> > > > > D.
> > > > >
> > > > > On Wed, Sep 15, 2021 at 8:46 PM Mike Drob wrote:
> > > > >>
> > > > >> The benchmark jar has the info we need… sort of. When I built it, it
> > > > >> has:
> > > > >>
> > > > >> Implementation-Version: 8.10.0
> > > > >> 75a5061d3715cc5d93c4cbe4f1fa62bf035eea1
> > > > >> 1 - mdrob - 2021-09-15 11:40:36
> > > > >>
> > > > >>
> > > > >> and it’s looking for Implementation-Version: 8.10.0
> > > > >> 75a5061d3715cc5d93c4cbe4f1fa62bf035eea11 on one line.
> > > > >>
> > > > >> Because 8.10 is a character longer than 8.9, we happen to wrap the
> > > > >> last character of the git commit sha. From the manifest spec:
> > > > >>
> > > > >> No line may be longer than 72 bytes (not characters), in its
> > > > >> UTF8-encoded form.
> > > > >> If a value would make the initial line longer than this, it should
> > > > >> be continued
> > > > >> on extra lines (each starting with a single SPACE).
> > > > >>
> > > > >> And we were already teetering on the edge of that limit. We'll run
> > > > >> into this problem again in a few years when we try to release
> > > > >> version 10.0.0, so solving it now has practical benefits down the
> > > > >> line.
> > > > >>
> > > > >> There's a few options that I can come up with -
> > > > >> 1. Use the short-hash when we generate the jar
> > > > >> 2. Use the short-hash when we check the contents in the smoke test
> > > > >> 3. Do some line join magic in the smoke test.
> > > > >>
> > > > >> I'm leaning towards number 1 as I feel that would still be unique
> > > > >> enough for our needs, but would
Re: New branch and feature freeze for Lucene/Solr 8.10.0
The shit that mike's python tool is doing, i'd expect bugs like that.
If you honestly disagree with me, go fix his python code to obey "the
rules". Be sure you have "the rules" correct as well (count bytes, not
characters, count EOL, etc).
Try to imagine how much software is going to get this wrong. Various
tools scanning for security problems that aren't written in java and
hence aren't using Jar class. Your security team isn't writing fucking
java code. Your ops team isn't writing fucking java code, so it really
does not matter if it is "easy in java".
This is *exactly* what i mean when i say this crazy shit has NO
INTEROP. It is 100% pure garbage, java showing its true colors.
On Thu, Sep 16, 2021 at 5:53 AM Robert Muir wrote:
>
> I was referring to doing this with languages other than java.
>
> I'm also assuming that exceeding this limit is going to cause indirect
> hassles for users of lucene, e.g. breaking various security / supply
> chain tools. We know a lot of these are total crap but people in the
> corporate world have to suffer under them.
>
> Its super-easy to use a short hash here and avoid problems.
>
> On Thu, Sep 16, 2021 at 3:03 AM Dawid Weiss wrote:
> >
> > Jar command doesn't have it, true. But it's fairly trivial to do, even
> > with an inline snippet like this one?
> >
> > public class PrintManifest {
> > public static void main(String[] jars) throws IOException {
> > for (var jar : jars) {
> > var manifest = new JarFile(Paths.get(jar).toFile()).getManifest();
> > var attrs = manifest.getMainAttributes();
> > System.out.println(jar + ": " +
> > attrs.getValue("Implementation-Version"));
> > }
> > }
> > }
> >
> > I have this in my lucene-core-9.0.0-SNAPSHOT.jar:
> >
> > Implementation-Version: 9.0.0-SNAPSHOT de45b68c909815ce5ea7b6b9e1a2ce337
> > 5b6334d [snapshot build, details omitted]
> >
> > and running:
> >
> > java PrintManifest.java lucene-core-9.0.0-SNAPSHOT.jar
> >
> > shows:
> >
> > lucene-core-9.0.0-SNAPSHOT.jar: 9.0.0-SNAPSHOT
> > de45b68c909815ce5ea7b6b9e1a2ce3375b6334d [snapshot build, details
> > omitted]
> >
> > This seems easier to me than trying to remember and keep the length of
> > that line shorter than an arbitrary limit.
> >
> > Dawid
> >
> >
> > On Wed, Sep 15, 2021 at 9:46 PM Robert Muir wrote:
> > >
> > > But its irrelevant that is "valid" when virtually no tools match it.
> > >
> > > In other words, I'd agree with you if the "jar" command had some
> > > ability to read these manifests and print stuff to stdout, e.g. if
> > > there was ANY interop at all here.
> > >
> > > But there isn't. So IMO it makes no sense to cause confusion and chaos
> > > by adding an unnecessarily long git commit hash.
> > >
> > > On Wed, Sep 15, 2021 at 3:26 PM Dawid Weiss wrote:
> > > >
> > > >
> > > > This is valid manifest line-breaking though... Can we read the manifest
> > > > properly on the smoke tester side somehow (for example, run a Java
> > > > process that reads and extracts the required attribute)? This way we
> > > > wouldn't care about the implementation details of how manifest wraps
> > > > the lines (or escapes characters).
> > > >
> > > > D.
> > > >
> > > > On Wed, Sep 15, 2021 at 8:46 PM Mike Drob wrote:
> > > >>
> > > >> The benchmark jar has the info we need… sort of. When I built it, it
> > > >> has:
> > > >>
> > > >> Implementation-Version: 8.10.0 75a5061d3715cc5d93c4cbe4f1fa62bf035eea1
> > > >> 1 - mdrob - 2021-09-15 11:40:36
> > > >>
> > > >>
> > > >> and it’s looking for Implementation-Version: 8.10.0
> > > >> 75a5061d3715cc5d93c4cbe4f1fa62bf035eea11 on one line.
> > > >>
> > > >> Because 8.10 is a character longer than 8.9, we happen to wrap the
> > > >> last character of the git commit sha. From the manifest spec:
> > > >>
> > > >> No line may be longer than 72 bytes (not characters), in its
> > > >> UTF8-encoded form.
> > > >> If a value would make the initial line longer than this, it should be
> > > >> continued
> > > >> on extra lines (each starting with a single SPACE).
> > > >>
> > > >> And we were already teetering on the edge of that limit. We'll run
> > > >> into this problem again in a few years when we try to release version
> > > >> 10.0.0, so solving it now has practical benefits down the line.
> > > >>
> > > >> There's a few options that I can come up with -
> > > >> 1. Use the short-hash when we generate the jar
> > > >> 2. Use the short-hash when we check the contents in the smoke test
> > > >> 3. Do some line join magic in the smoke test.
> > > >>
> > > >> I'm leaning towards number 1 as I feel that would still be unique
> > > >> enough for our needs, but would like to hear from others as well.
> > > >>
> > > >> On Wed, Sep 15, 2021 at 9:46 AM Timothy potter
> > > >> wrote:
> > > >>>
> > > >>> can someone also please look into that benchmark jar issue?
> > > >>>
> > > >>> Sent from my iPhone
> > > >>>
> > > >>> On Sep 15, 2021, at 9:44 AM, Nhat Nguyen
> > > >>> wrote:
> > > >>>
> >
Re: New branch and feature freeze for Lucene/Solr 8.10.0
I was referring to doing this with languages other than java.
I'm also assuming that exceeding this limit is going to cause indirect
hassles for users of lucene, e.g. breaking various security / supply
chain tools. We know a lot of these are total crap but people in the
corporate world have to suffer under them.
Its super-easy to use a short hash here and avoid problems.
On Thu, Sep 16, 2021 at 3:03 AM Dawid Weiss wrote:
>
> Jar command doesn't have it, true. But it's fairly trivial to do, even
> with an inline snippet like this one?
>
> public class PrintManifest {
> public static void main(String[] jars) throws IOException {
> for (var jar : jars) {
> var manifest = new JarFile(Paths.get(jar).toFile()).getManifest();
> var attrs = manifest.getMainAttributes();
> System.out.println(jar + ": " +
> attrs.getValue("Implementation-Version"));
> }
> }
> }
>
> I have this in my lucene-core-9.0.0-SNAPSHOT.jar:
>
> Implementation-Version: 9.0.0-SNAPSHOT de45b68c909815ce5ea7b6b9e1a2ce337
> 5b6334d [snapshot build, details omitted]
>
> and running:
>
> java PrintManifest.java lucene-core-9.0.0-SNAPSHOT.jar
>
> shows:
>
> lucene-core-9.0.0-SNAPSHOT.jar: 9.0.0-SNAPSHOT
> de45b68c909815ce5ea7b6b9e1a2ce3375b6334d [snapshot build, details
> omitted]
>
> This seems easier to me than trying to remember and keep the length of
> that line shorter than an arbitrary limit.
>
> Dawid
>
>
> On Wed, Sep 15, 2021 at 9:46 PM Robert Muir wrote:
> >
> > But its irrelevant that is "valid" when virtually no tools match it.
> >
> > In other words, I'd agree with you if the "jar" command had some
> > ability to read these manifests and print stuff to stdout, e.g. if
> > there was ANY interop at all here.
> >
> > But there isn't. So IMO it makes no sense to cause confusion and chaos
> > by adding an unnecessarily long git commit hash.
> >
> > On Wed, Sep 15, 2021 at 3:26 PM Dawid Weiss wrote:
> > >
> > >
> > > This is valid manifest line-breaking though... Can we read the manifest
> > > properly on the smoke tester side somehow (for example, run a Java
> > > process that reads and extracts the required attribute)? This way we
> > > wouldn't care about the implementation details of how manifest wraps the
> > > lines (or escapes characters).
> > >
> > > D.
> > >
> > > On Wed, Sep 15, 2021 at 8:46 PM Mike Drob wrote:
> > >>
> > >> The benchmark jar has the info we need… sort of. When I built it, it has:
> > >>
> > >> Implementation-Version: 8.10.0 75a5061d3715cc5d93c4cbe4f1fa62bf035eea1
> > >> 1 - mdrob - 2021-09-15 11:40:36
> > >>
> > >>
> > >> and it’s looking for Implementation-Version: 8.10.0
> > >> 75a5061d3715cc5d93c4cbe4f1fa62bf035eea11 on one line.
> > >>
> > >> Because 8.10 is a character longer than 8.9, we happen to wrap the last
> > >> character of the git commit sha. From the manifest spec:
> > >>
> > >> No line may be longer than 72 bytes (not characters), in its
> > >> UTF8-encoded form.
> > >> If a value would make the initial line longer than this, it should be
> > >> continued
> > >> on extra lines (each starting with a single SPACE).
> > >>
> > >> And we were already teetering on the edge of that limit. We'll run into
> > >> this problem again in a few years when we try to release version 10.0.0,
> > >> so solving it now has practical benefits down the line.
> > >>
> > >> There's a few options that I can come up with -
> > >> 1. Use the short-hash when we generate the jar
> > >> 2. Use the short-hash when we check the contents in the smoke test
> > >> 3. Do some line join magic in the smoke test.
> > >>
> > >> I'm leaning towards number 1 as I feel that would still be unique enough
> > >> for our needs, but would like to hear from others as well.
> > >>
> > >> On Wed, Sep 15, 2021 at 9:46 AM Timothy potter
> > >> wrote:
> > >>>
> > >>> can someone also please look into that benchmark jar issue?
> > >>>
> > >>> Sent from my iPhone
> > >>>
> > >>> On Sep 15, 2021, at 9:44 AM, Nhat Nguyen
> > >>> wrote:
> > >>>
> > >>>
> > >>> Thanks Mayya and Mike! I will backport it to the 8.10 branch.
> > >>>
> > >>> On Wed, Sep 15, 2021 at 10:12 AM Mike Drob wrote:
> >
> > I think since Tim is out on vacation, it's probably not too late. That
> > looks like a good fix to have, do we know how long the bug has been
> > present?
> >
> > On Wed, Sep 15, 2021 at 7:56 AM Mayya Sharipova
> > wrote:
> > >
> > > Hello everyone,
> > > We have discovered a bug and fixed a bug in Lucene sort optimization
> > > (LUCENE-10106) and would like to merge it to Lucene 8.10 if it is not
> > > too late.
> > > I apologize for the inconvenience, the bug was discovered just
> > > yesterday.
> > >
> > > On Tue, Sep 14, 2021 at 9:26 PM Timothy Potter
> > > wrote:
> > >>
> > >> Ahem ... unfortunately there will not be an 8.10 RC this week. I'm
> > >> headed out on vacation tomorrow, back at keys on Monday, Sept
[jira] [Created] (PYLUCENE-60) we are providing the projects for mtech and btech students.
Nasima Takeoff created PYLUCENE-60: -- Summary: we are providing the projects for mtech and btech students. Key: PYLUCENE-60 URL: https://issues.apache.org/jira/browse/PYLUCENE-60 Project: PyLucene Issue Type: Blog - New Blog Request Environment: python Reporter: Nasima Takeoff Attachments: Python projects.png h4. Python has been within the top 10 popular programming languages for an extended time because the community of Python programmers has grown tons thanks to its easy syntax and library support. during this article, Takeoff will be able to introduce you to 60 amazing Python projects with ASCII text files solved and explained by experts in [Takeoff Projects|https://takeoffprojects.com/]. -- This message was sent by Atlassian Jira (v8.3.4#803005)
Re: New branch and feature freeze for Lucene/Solr 8.10.0
Jar command doesn't have it, true. But it's fairly trivial to do, even
with an inline snippet like this one?
public class PrintManifest {
public static void main(String[] jars) throws IOException {
for (var jar : jars) {
var manifest = new JarFile(Paths.get(jar).toFile()).getManifest();
var attrs = manifest.getMainAttributes();
System.out.println(jar + ": " + attrs.getValue("Implementation-Version"));
}
}
}
I have this in my lucene-core-9.0.0-SNAPSHOT.jar:
Implementation-Version: 9.0.0-SNAPSHOT de45b68c909815ce5ea7b6b9e1a2ce337
5b6334d [snapshot build, details omitted]
and running:
java PrintManifest.java lucene-core-9.0.0-SNAPSHOT.jar
shows:
lucene-core-9.0.0-SNAPSHOT.jar: 9.0.0-SNAPSHOT
de45b68c909815ce5ea7b6b9e1a2ce3375b6334d [snapshot build, details
omitted]
This seems easier to me than trying to remember and keep the length of
that line shorter than an arbitrary limit.
Dawid
On Wed, Sep 15, 2021 at 9:46 PM Robert Muir wrote:
>
> But its irrelevant that is "valid" when virtually no tools match it.
>
> In other words, I'd agree with you if the "jar" command had some
> ability to read these manifests and print stuff to stdout, e.g. if
> there was ANY interop at all here.
>
> But there isn't. So IMO it makes no sense to cause confusion and chaos
> by adding an unnecessarily long git commit hash.
>
> On Wed, Sep 15, 2021 at 3:26 PM Dawid Weiss wrote:
> >
> >
> > This is valid manifest line-breaking though... Can we read the manifest
> > properly on the smoke tester side somehow (for example, run a Java process
> > that reads and extracts the required attribute)? This way we wouldn't care
> > about the implementation details of how manifest wraps the lines (or
> > escapes characters).
> >
> > D.
> >
> > On Wed, Sep 15, 2021 at 8:46 PM Mike Drob wrote:
> >>
> >> The benchmark jar has the info we need… sort of. When I built it, it has:
> >>
> >> Implementation-Version: 8.10.0 75a5061d3715cc5d93c4cbe4f1fa62bf035eea1
> >> 1 - mdrob - 2021-09-15 11:40:36
> >>
> >>
> >> and it’s looking for Implementation-Version: 8.10.0
> >> 75a5061d3715cc5d93c4cbe4f1fa62bf035eea11 on one line.
> >>
> >> Because 8.10 is a character longer than 8.9, we happen to wrap the last
> >> character of the git commit sha. From the manifest spec:
> >>
> >> No line may be longer than 72 bytes (not characters), in its UTF8-encoded
> >> form.
> >> If a value would make the initial line longer than this, it should be
> >> continued
> >> on extra lines (each starting with a single SPACE).
> >>
> >> And we were already teetering on the edge of that limit. We'll run into
> >> this problem again in a few years when we try to release version 10.0.0,
> >> so solving it now has practical benefits down the line.
> >>
> >> There's a few options that I can come up with -
> >> 1. Use the short-hash when we generate the jar
> >> 2. Use the short-hash when we check the contents in the smoke test
> >> 3. Do some line join magic in the smoke test.
> >>
> >> I'm leaning towards number 1 as I feel that would still be unique enough
> >> for our needs, but would like to hear from others as well.
> >>
> >> On Wed, Sep 15, 2021 at 9:46 AM Timothy potter
> >> wrote:
> >>>
> >>> can someone also please look into that benchmark jar issue?
> >>>
> >>> Sent from my iPhone
> >>>
> >>> On Sep 15, 2021, at 9:44 AM, Nhat Nguyen
> >>> wrote:
> >>>
> >>>
> >>> Thanks Mayya and Mike! I will backport it to the 8.10 branch.
> >>>
> >>> On Wed, Sep 15, 2021 at 10:12 AM Mike Drob wrote:
>
> I think since Tim is out on vacation, it's probably not too late. That
> looks like a good fix to have, do we know how long the bug has been
> present?
>
> On Wed, Sep 15, 2021 at 7:56 AM Mayya Sharipova
> wrote:
> >
> > Hello everyone,
> > We have discovered a bug and fixed a bug in Lucene sort optimization
> > (LUCENE-10106) and would like to merge it to Lucene 8.10 if it is not
> > too late.
> > I apologize for the inconvenience, the bug was discovered just
> > yesterday.
> >
> > On Tue, Sep 14, 2021 at 9:26 PM Timothy Potter
> > wrote:
> >>
> >> Ahem ... unfortunately there will not be an 8.10 RC this week. I'm
> >> headed out on vacation tomorrow, back at keys on Monday, Sept 20
> >> unless someone else wants to pick up the RM duties before then?
> >>
> >> After failing the test suite at various places and other weirdness
> >> like .asc files not getting created, I finally got to the smoke test
> >> part, which is now failing with:
> >>
> >> File
> >> "/Users/tjp/.lucene-releases/8.10.0/lucene-solr/dev-tools/scripts/smokeTestRelease.py",
> >> line 176, in checkJARMetaData
> >> raise RuntimeError('%s is missing "%s" inside its
> >> META-INF/MANIFEST.MF (wrong git revision?)' % \
> >> RuntimeError: JAR file
> >>
