Skip to contentSkip to conversation

Add sequence number to 'xm info'

PATCH In review Dan Smith 3 participants 6 messages Sep 28, 2005 – Sep 29, 2005

lore
Share & tools
lore

Dan Smith

author
#0 lore raw

PATCH 1/1 Add sequence number to 'xm info'

This patch causes the short changeset id or sequence number to be included on the 'xen_changeset' line in the 'xm info' output.

We plan to use this sequence number as a very fuzzy way of sorting test reports so that we can start looking for trends. While we know it won't result in a perfect ordering of changesets for reporting, it will help us get close.

xen/Makefile +1 -1
diff -r 5e5ae8340956 xen/Makefile
@@ -99,7 +99,7 @@
 	    -e 's/@@version@@/$(XEN_VERSION)/g' \
 	    -e 's/@@subversion@@/$(XEN_SUBVERSION)/g' \
 	    -e 's/@@extraversion@@/$(XEN_EXTRAVERSION)/g' \
-	    -e 's!@@changeset@@!$(shell (hg parents | awk -F: '/^changeset/{CS=$$3};{FS="date:[ ]+"}/^date/{D=$$2}; END {print D, CS}') 2>/dev/null || (head -n 6 ChangeLog | awk -F: '/^changeset/{CS=$$3};{FS="date:[ ]+"}/^date/{D=$$2}; END {print D, CS}') 2>/dev/null || echo information unavailable)!g' \
+	    -e 's!@@changeset@@!$(shell (hg parents | awk '{FS="changeset:[ ]+"}/^changeset/{CS=$$2};{FS="date:[ ]+"}/^date/{D=$$2}; END {print D, CS}') 2>/dev/null || (head -n 6 ChangeLog | awk -F: '/^changeset/{CS=$$3};{FS="date:[ ]+"}/^date/{D=$$2}; END {print D, CS}') 2>/dev/null || echo information unavailable)!g' \
 	    < include/xen/compile.h.in > [email protected]
 	@cat include/xen/banner.h >> [email protected]
 	@mv -f [email protected] $@
Signature · 4 lines
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: [email protected]

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel

#1 lore raw
Quoted 7 lines · Dan Smith

This patch causes the short changeset id or sequence number to be included on the 'xen_changeset' line in the 'xm info' output.

We plan to use this sequence number as a very fuzzy way of sorting test reports so that we can start looking for trends. While we know it won't result in a perfect ordering of changesets for reporting, it will help us get close.

So if the new format is

	xen_changeset          : Wed Sep 28 13:06:49 2005 +0100 

7120:5e5ae8340956
This is a bad idea. If you want to compare two changesets, you already have the full date right in front of you!

The revision number is a convenience for *local* operations, for example 'hg export 7033:7035'. It obviously should never be compared across different repositories "and just hope things work out ok."

Signature · 2 lines
Hollis Blanchard
IBM Linux Technology Center
#2 lore raw
Quoted 11 lines · Hollis Blanchard
Quoted 7 lines · Dan Smith

This patch causes the short changeset id or sequence number to be included on the 'xen_changeset' line in the 'xm info' output.

We plan to use this sequence number as a very fuzzy way of sorting test reports so that we can start looking for trends. While we know it won't result in a perfect ordering of changesets for reporting, it will help us get close.

So if the new format is

Quoted 8 lines
	xen_changeset          : Wed Sep 28 13:06:49 2005 +0100 

7120:5e5ae8340956
This is a bad idea. If you want to compare two changesets, you already have the full date right in front of you!

The revision number is a convenience for *local* operations, for example 'hg export 7033:7035'. It obviously should never be compared across different repositories "and just hope things work out ok."

Yes, single repository, like xen-unstable (upstream). For the xm test work having a sequence number from a single repo makes a number of things much easier. If it was so harmful to have the sequence number then I would think it would be eliminated from hg entirely.

Most people using hg for xen-unstable aren't doing so to write code, they are doing so to make it easy to test latest pull (akin to anon cvs). For that class of people and applications, having sequence number available is a very good thing.

-Sean

Signature · 9 lines
__________________________________________________________________

Sean Dague                                       Mid-Hudson Valley
sean at dague dot net                            Linux Users Group
http://dague.net                                 http://mhvlug.org

There is no silver bullet.  Plus, werewolves make better neighbors
than zombies, and they tend to keep the vampire population down.
__________________________________________________________________

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel

author
#4 lore raw
In reply to Hollis Blanchard

HB> +If you want to compare two changesets, you already have the full HB> +date right in front of you!

That's true.

HB> The revision number is a convenience for *local* operations, for HB> example 'hg export 7033:7035'. It obviously should never be HB> compared across different repositories "and just hope things work HB> out ok."

I completely agree and understand. Telling someone to try to resolve their problem by checking out changeset 7033 would be a bad idea. However, if we're talking about loosely grouping and sorting a month's worth of test reports to make a determination about failure trends, I think it's a valid way to do it. It's quick and it doesn't require any parsing of the date string.

Also, when I'm comparing two of my test machines to see which is running a newer pull, I have to parse the date string with my eyes and do timezone conversions to figure out the order. Since my (and most, I imagine) test machines are always running clones of the main repo, the sequence numbers would always be valid.

Independent of how people choose to use the information, is there a strong argument for not even showing it?

Signature · 4 lines
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: [email protected]
#3 lore raw
In reply to Sean Dague
Quoted 16 lines · Sean Dague
Quoted 15 lines · Hollis Blanchard
Quoted 7 lines · Dan Smith

This patch causes the short changeset id or sequence number to be included on the 'xen_changeset' line in the 'xm info' output.

We plan to use this sequence number as a very fuzzy way of sorting test reports so that we can start looking for trends. While we know it won't result in a perfect ordering of changesets for reporting, it will help us get close.

So if the new format is

	xen_changeset          : Wed Sep 28 13:06:49 2005 +0100

7120:5e5ae8340956
This is a bad idea. If you want to compare two changesets, you already have the full date right in front of you!

Quoted 8 lines
Quoted 3 lines

The revision number is a convenience for *local* operations, for example 'hg export 7033:7035'. It obviously should never be compared across different repositories "and just hope things work out ok."

Yes, single repository, like xen-unstable (upstream). For the xm test work having a sequence number from a single repo makes a number of things much easier. If it was so harmful to have the sequence number then I would think it would be eliminated from hg entirely.

The revision number is useful, as you can see in my export example above -- there would be no easy way to specify a range of changesets without it.

Quoted 4 lines

Most people using hg for xen-unstable aren't doing so to write code, they are doing so to make it easy to test latest pull (akin to anon cvs). For that class of people and applications, having sequence number available is a very good thing.

I haven't seen Dan's intended use case described, but there are better sources of information, such as the changeset hash or the changeset date.

Signature · 2 lines
Hollis Blanchard
IBM Linux Technology Center
#5 lore raw
In reply to Dan Smith
Quoted 10 lines · Dan Smith

HB> +If you want to compare two changesets, you already have the full HB> +date right in front of you!

That's true.

HB> The revision number is a convenience for *local* operations, for HB> example 'hg export 7033:7035'. It obviously should never be HB> compared across different repositories "and just hope things work HB> out ok."

Quoted 6 lines

I completely agree and understand. Telling someone to try to resolve their problem by checking out changeset 7033 would be a bad idea. However, if we're talking about loosely grouping and sorting a month's worth of test reports to make a determination about failure trends, I think it's a valid way to do it. It's quick and it doesn't require any parsing of the date string.

Look at the way you phrased that: you aren't saying "a hundred revisions"; you're saying "a month". Maybe you could describe this classification process a little better. Are you looking to make statements like "dom0 booting was broken every week"?

And let's be honest: I'm not a big math fan, but the arithmetic on time zone conversions is not very difficult. :)

Quoted 5 lines

Also, when I'm comparing two of my test machines to see which is running a newer pull, I have to parse the date string with my eyes and do timezone conversions to figure out the order. Since my (and most, I imagine) test machines are always running clones of the main repo, the sequence numbers would always be valid.

Sure, although if someone forgets what tree they're currently working in (happens to me anyways :), or doesn't have a pristine tree handy, or whatever...

Quoted 2 lines

Independent of how people choose to use the information, is there a strong argument for not even showing it?

Because if you show it then people might be tempted to use the information. :) Revision numbers are local numbers, and should only be used locally.

I don't think "in this one case it will usually be ok" is a good argument, especially since "will always work" methods are readily available.

Signature · 2 lines
Hollis Blanchard
IBM Linux Technology Center