Обсуждение: [pgsql-www] buildfarm git references pointing at github?
If you go to, say, https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=koreaceratops&dt=2017-03-13%2020%3A49%3A29 and check the git links near the bottom of the page, you'll notice they all point at github, not at git.postgresql.org. I do not like that, first because I do not consider github to be an authoritative source for this purpose, and second because the links are all 404 anyway. Some other nearby pages have only some of the links pointing at github, but why are any of them doing that? regards, tom lane
On 03/13/2017 11:06 PM, Tom Lane wrote: > If you go to, say, > > https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=koreaceratops&dt=2017-03-13%2020%3A49%3A29 > > and check the git links near the bottom of the page, you'll notice they > all point at github, not at git.postgresql.org. I do not like that, > first because I do not consider github to be an authoritative source > for this purpose, and second because the links are all 404 anyway. > > Some other nearby pages have only some of the links pointing at github, > but why are any of them doing that? > > This actually goes back to the time when we were switching from CVS to git. We can clean that up now, I think :-) As for use of github, its use is hardcoded in the case of multi-commit changesets because I don't know a way to make the standard gitweb display those, although it might exist. And TBH github's display of indiviual commits is lots nicer, too. Compare <https://github.com/postgres/postgres/commit/a82178020de2c4b3576d0a079e11a431e382e4de> with <https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=a82178020de2c4b3576d0a079e11a431e382e4de> The github repo is a mirror of the repo at git.postgresql.org. I don't see a problem with using it. cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
> On 14 Mar 2017, at 13:53, Andrew Dunstan <andrew.dunstan@2ndquadrant.com> wrote: <snip> > ... github's display of indiviual commits is lots nicer, too. Compare > <https://github.com/postgres/postgres/commit/a82178020de2c4b3576d0a079e11a431e382e4de> > with > <https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=a82178020de2c4b3576d0a079e11a431e382e4de> As a data point, the "commitdiff" tab instead of the "commit" one seems closer in nature to the GH one: https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=a82178020de2c4b3576d0a079e11a431e382e4de + Justin -- "My grandfather once told me that there are two kinds of people: those who work and those who take the credit. He told me to try to be in the first group; there was less competition there." - Indira Gandhi
Andrew Dunstan <andrew.dunstan@2ndquadrant.com> writes: > As for use of github, its use is hardcoded in the case of multi-commit > changesets because I don't know a way to make the standard gitweb > display those, although it might exist. And TBH github's display of > indiviual commits is lots nicer, too. Compare > <https://github.com/postgres/postgres/commit/a82178020de2c4b3576d0a079e11a431e382e4de> > with > <https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=a82178020de2c4b3576d0a079e11a431e382e4de> Beauty is in the eye of the beholder I guess, but to me the first one is a prime example of crappy webdesign --- too busy, too convinced that randomly colored backgrounds and legibility-killing font coloring choices are great ideas. I really would rather not be directed to github when checking on our commits. regards, tom lane
On 03/14/2017 10:15 AM, Tom Lane wrote: > Andrew Dunstan <andrew.dunstan@2ndquadrant.com> writes: >> As for use of github, its use is hardcoded in the case of multi-commit >> changesets because I don't know a way to make the standard gitweb >> display those, although it might exist. And TBH github's display of >> indiviual commits is lots nicer, too. Compare >> <https://github.com/postgres/postgres/commit/a82178020de2c4b3576d0a079e11a431e382e4de> >> with >> <https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=a82178020de2c4b3576d0a079e11a431e382e4de> > Beauty is in the eye of the beholder I guess, but to me the first one > is a prime example of crappy webdesign --- too busy, too convinced that > randomly colored backgrounds and legibility-killing font coloring choices > are great ideas. So I guess you're also not keen on gitweb's commitdiff display suggested nearby by Justin Clift? Is there anyone else who wants to stick with the commit display instead of switching to commitdiff? > I really would rather not be directed to github when > checking on our commits. I have already changed that except for the case where gitweb provides no equivalent functionality (the "Change set" links). cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
Andrew Dunstan <andrew.dunstan@2ndquadrant.com> writes:
> On 03/14/2017 10:15 AM, Tom Lane wrote:
>> Beauty is in the eye of the beholder I guess, but to me the first one
>> is a prime example of crappy webdesign --- too busy, too convinced that
>> randomly colored backgrounds and legibility-killing font coloring choices
>> are great ideas.
> So I guess you're also not keen on gitweb's commitdiff display suggested
> nearby by Justin Clift?
Well, I find it a considerable improvement over github's, at least in this
example.
> Is there anyone else who wants to stick with the commit display instead
> of switching to commitdiff?
+1 for commitdiff, that's pretty much always the form of the page I want
to see.
regards, tom lane
On 03/14/2017 10:32 AM, Tom Lane wrote: > Andrew Dunstan <andrew.dunstan@2ndquadrant.com> writes: >> On 03/14/2017 10:15 AM, Tom Lane wrote: >>> Beauty is in the eye of the beholder I guess, but to me the first one >>> is a prime example of crappy webdesign --- too busy, too convinced that >>> randomly colored backgrounds and legibility-killing font coloring choices >>> are great ideas. >> So I guess you're also not keen on gitweb's commitdiff display suggested >> nearby by Justin Clift? > Well, I find it a considerable improvement over github's, at least in this > example. > >> Is there anyone else who wants to stick with the commit display instead >> of switching to commitdiff? > +1 for commitdiff, that's pretty much always the form of the page I want > to see. OK, We'll do that. cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
On 03/14/2017 10:32 AM, Tom Lane wrote: > > +1 for commitdiff, that's pretty much always the form of the page I want > to see. Done. cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
Andrew Dunstan wrote: > On 03/14/2017 10:15 AM, Tom Lane wrote: > > I really would rather not be directed to github when > > checking on our commits. > > I have already changed that except for the case where gitweb provides no > equivalent functionality (the "Change set" links). https://git.postgresql.org/gitweb/?p=postgresql.git;a=shortlog;hp=9e09264;h=9ca5c8721dfa -- Álvaro Herrera https://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
On 03/14/2017 10:45 AM, Alvaro Herrera wrote: > Andrew Dunstan wrote: > >> On 03/14/2017 10:15 AM, Tom Lane wrote: >>> I really would rather not be directed to github when >>> checking on our commits. >> I have already changed that except for the case where gitweb provides no >> equivalent functionality (the "Change set" links). > https://git.postgresql.org/gitweb/?p=postgresql.git;a=shortlog;hp=9e09264;h=9ca5c8721dfa > Yes, but you can't get the diff for the whole changeset AFAIK. If you click the commitdiff button for that link it only shows you the diff for one commit. And we're already showing you the summary in the buildfarm web page, so there's not much value in changing it to this. cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
Andrew Dunstan <andrew.dunstan@2ndquadrant.com> writes: > On 03/14/2017 10:32 AM, Tom Lane wrote: >> +1 for commitdiff, that's pretty much always the form of the page I want >> to see. > Done. Thanks for changing that, it's definitely helpful. I'm a bit confused about the links for the "files changed this run" list. What they seem to be now is pointers to the last commit that changed each file. I might be misremembering, but I thought they had been links to the individual files in our git tree. As long as we're fooling around here, it seems like links to the per-file history logs would be convenient; eg instead of http://git.postgresql.org/gitweb?p=postgresql.git;a=commitdiff;h=a82178020de2c4b3576d0a079e11a431e382e4de which is the link being offered right now for src/backend/optimizer/plan/createplan.c in recent runs, perhaps https://git.postgresql.org/gitweb/?p=postgresql.git;a=history;f=src/backend/optimizer/plan/createplan.c;h=4629ca27ee559086929daf437ac5e6954b1987d6 That particular display is useful since it's easy to get to either the raw file contents ("blob") or the relevant commitdiffs from there. However, fooling with it, it seems that the blob links don't get offered unless the "h=" hash is the file's hash, not the commit's hash. (gitweb bug?) That might make it too hard to do, dunno. A close second in usefulness would be to use the commit's hash, https://git.postgresql.org/gitweb/?p=postgresql.git;a=history;f=src/backend/optimizer/plan/createplan.c;h=a82178020de2c4b3576d0a079e11a431e382e4de regards, tom lane
Andrew Dunstan wrote: > > > On 03/14/2017 10:45 AM, Alvaro Herrera wrote: > > Andrew Dunstan wrote: > > > >> On 03/14/2017 10:15 AM, Tom Lane wrote: > >>> I really would rather not be directed to github when > >>> checking on our commits. > >> I have already changed that except for the case where gitweb provides no > >> equivalent functionality (the "Change set" links). > > https://git.postgresql.org/gitweb/?p=postgresql.git;a=shortlog;hp=9e09264;h=9ca5c8721dfa > > Yes, but you can't get the diff for the whole changeset AFAIK. If you > click the commitdiff button for that link it only shows you the diff for > one commit. And we're already showing you the summary in the buildfarm > web page, so there's not much value in changing it to this. Ah, so what you want is this: https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;hp=1e6de941e;h=9ca5c8721dfa (I changed the "hp" parent commit to avoid a large diff). -- Álvaro Herrera https://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
On 03/14/2017 11:14 AM, Alvaro Herrera wrote: > Andrew Dunstan wrote: >> >> On 03/14/2017 10:45 AM, Alvaro Herrera wrote: >>> Andrew Dunstan wrote: >>> >>>> On 03/14/2017 10:15 AM, Tom Lane wrote: >>>>> I really would rather not be directed to github when >>>>> checking on our commits. >>>> I have already changed that except for the case where gitweb provides no >>>> equivalent functionality (the "Change set" links). >>> https://git.postgresql.org/gitweb/?p=postgresql.git;a=shortlog;hp=9e09264;h=9ca5c8721dfa >> Yes, but you can't get the diff for the whole changeset AFAIK. If you >> click the commitdiff button for that link it only shows you the diff for >> one commit. And we're already showing you the summary in the buildfarm >> web page, so there's not much value in changing it to this. > Ah, so what you want is this: > https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;hp=1e6de941e;h=9ca5c8721dfa > (I changed the "hp" parent commit to avoid a large diff). > That is still only showing me the diff for a single commit AFAICT. cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
Andrew Dunstan wrote: > > Ah, so what you want is this: > > https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;hp=1e6de941e;h=9ca5c8721dfa > > (I changed the "hp" parent commit to avoid a large diff). > > That is still only showing me the diff for a single commit AFAICT. No, it's the diff for all commits in the interval. I admit that the fact that the header shows only one commit invites confusion -- gitweb bug, clearly. Compare: https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=9ca5c8721dfa -- Álvaro Herrera https://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
On 03/14/2017 11:29 AM, Alvaro Herrera wrote: > Andrew Dunstan wrote: > >>> Ah, so what you want is this: >>> https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;hp=1e6de941e;h=9ca5c8721dfa >>> (I changed the "hp" parent commit to avoid a large diff). >> That is still only showing me the diff for a single commit AFAICT. > No, it's the diff for all commits in the interval. I admit that the > fact that the header shows only one commit invites confusion -- gitweb > bug, clearly. Compare: > https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=9ca5c8721dfa > Yeah, That's very, very confusing. I'll be happy to change it if that bug gets fixed. cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
On 03/14/2017 11:07 AM, Tom Lane wrote: > Andrew Dunstan <andrew.dunstan@2ndquadrant.com> writes: >> On 03/14/2017 10:32 AM, Tom Lane wrote: >>> +1 for commitdiff, that's pretty much always the form of the page I want >>> to see. >> Done. > Thanks for changing that, it's definitely helpful. > > I'm a bit confused about the links for the "files changed this run" > list. What they seem to be now is pointers to the last commit that > changed each file. I might be misremembering, but I thought they > had been links to the individual files in our git tree. As long > as we're fooling around here, it seems like links to the per-file > history logs would be convenient; eg instead of > > http://git.postgresql.org/gitweb?p=postgresql.git;a=commitdiff;h=a82178020de2c4b3576d0a079e11a431e382e4de > > which is the link being offered right now for > src/backend/optimizer/plan/createplan.c in recent runs, perhaps > > https://git.postgresql.org/gitweb/?p=postgresql.git;a=history;f=src/backend/optimizer/plan/createplan.c;h=4629ca27ee559086929daf437ac5e6954b1987d6 > > That particular display is useful since it's easy to get to either > the raw file contents ("blob") or the relevant commitdiffs from there. > However, fooling with it, it seems that the blob links don't get offered > unless the "h=" hash is the file's hash, not the commit's hash. (gitweb > bug?) That might make it too hard to do, dunno. A close second in > usefulness would be to use the commit's hash, > > https://git.postgresql.org/gitweb/?p=postgresql.git;a=history;f=src/backend/optimizer/plan/createplan.c;h=a82178020de2c4b3576d0a079e11a431e382e4de > > It's pretty much always been this way, pointing to the most recent relevant commit. Originally, the server knew nothing of git refs (or CVS revision numbers) at all. It just took information from the client about versions of each file and shoved it into a URL, and the client code has always pointed to the most recent revision in the relevant time period. We could do a lot better than that now, as we know the git commit ref for each snapshot, and the server keeps a pretty up to data mirror of the code. I agree it would be nice to get a history for the file for the relevant period. I didn't really like either of your suggestions, though, because they are showing a lot more history that I want to see. However, there's more work there than just a line or two to change, so it's going to take a bit of time to get right. cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services