Обсуждение: [HACKERS] show precise repos version for dev builds?
Hello,
I use postgres "11devel" packages kindly distributed on
"apt.postgresql.org" and recompiled every few hours.
I wanted to know which version it was, and "11devel" is kind of imprecise.
Ok, there is a hint in the deb package name:
11~~devel~20170930.2214-1~87.git2632bcc.pgdg16.04+1
But this information does not seem to be available from the executables
themselves:
sh> psql --version psql (PostgreSQL) 11devel
sh> pgbench --version pgbench (PostgreSQL) 9.6.5
Some projects provide a repository or build date information:
sh> clang --version clang version 6.0.0 (trunk 314585)
sh> gcc --version gcc (GCC) 8.0.0 20170930 (experimental)
With some svn project I use "svnversion" which shows a summary of the
status, eg "5432M" which tells that the sources are locally modified
from version 5432.
I would find it useful to have something like that in pg as well, and I
have not found it available.
ISTM that extending the version name with the commit id and or date in
some version output, eg "11devel [2632bcc 2017-09-30 ...]", would do it.
Thoughts?
--
Fabien.
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
Fabien COELHO <coelho@cri.ensmp.fr> writes:
> I wanted to know which version it was, and "11devel" is kind of imprecise.
> ...
> ISTM that extending the version name with the commit id and or date in
> some version output, eg "11devel [2632bcc 2017-09-30 ...]", would do it.
configure --with-extra-version=whateveryouwant
regards, tom lane
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
Hello, > Fabien COELHO <coelho@cri.ensmp.fr> writes: >> I wanted to know which version it was, and "11devel" is kind of imprecise. >> ... >> ISTM that extending the version name with the commit id and or date in >> some version output, eg "11devel [2632bcc 2017-09-30 ...]", would do it. > > configure --with-extra-version=whateveryouwant Thanks for the pointer! So now I have to convince the apt.postgresql.org people to build the devel version with this trick. -- Fabien. -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
On Sun, Oct 1, 2017 at 8:10 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote: > > configure --with-extra-version=whateveryouwant I see that this build option has been around since 9.4; is anyone using it to mark patched production builds? EnterpriseDB or 2ndQuadrant? How about the cloud providers? -Jeremy -- http://about.me/jeremy_schneider -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
On 11 October 2017 at 11:44, Jeremy Schneider <schneider@ardentperf.com> wrote: > On Sun, Oct 1, 2017 at 8:10 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote: >> >> configure --with-extra-version=whateveryouwant > > I see that this build option has been around since 9.4; is anyone > using it to mark patched production builds? EnterpriseDB or > 2ndQuadrant? How about the cloud providers? We started using it for BDR, but unfortunately too much software explodes spectacularly when you use it, due to simplistic/buggy version parsing. This led me to push for wider adoption of server_version_num - in particular, exposing it as an initial connection parameter via GUC_REPORT, and exposing it in pg_config. After a few attempts I've given up on that as a lost cause now, though I still disagree with the rationale. Right now, if you use it, various 3rd party software will fail in a variety of exciting ways, some more subtle than others. -- Craig Ringer http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
On Wed, Oct 11, 2017 at 1:19 AM, Craig Ringer <craig@2ndquadrant.com> wrote: > We started using it for BDR, but unfortunately too much software > explodes spectacularly when you use it, due to simplistic/buggy > version parsing. Since 10.0 will break most of that software anyway, maybe this is a good time to revisit the question? :) -J -- http://about.me/jeremy_schneider -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
On 10/11/17 04:19, Craig Ringer wrote: > On 11 October 2017 at 11:44, Jeremy Schneider <schneider@ardentperf.com> wrote: >> On Sun, Oct 1, 2017 at 8:10 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote: >>> >>> configure --with-extra-version=whateveryouwant >> >> I see that this build option has been around since 9.4; is anyone >> using it to mark patched production builds? EnterpriseDB or >> 2ndQuadrant? How about the cloud providers? > > We started using it for BDR, but unfortunately too much software > explodes spectacularly when you use it, due to simplistic/buggy > version parsing. I've been using --with-extra-version=+git`date +%Y%m%d`"~"`git rev-parse --short HEAD` for my local builds for some time, and I've not experienced any such problems. However, using the various numeric reporting options is clearly better if you want to do version comparisons. The "extra version" stuff should be mainly for labeling. -- Peter Eisentraut http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
On 12 October 2017 at 06:46, Peter Eisentraut <peter.eisentraut@2ndquadrant.com> wrote: > I've been using > > --with-extra-version=+git`date +%Y%m%d`"~"`git rev-parse --short HEAD` > > for my local builds for some time, and I've not experienced any such > problems. Interesting. I've seen issues with a number of tools. The one I can remember most clearly is check_postgres.pl . Nobody's going to argue that this is pretty code, but last time I tested (9.4-era, admittedly) it exploded messily with extra-version. > However, using the various numeric reporting options is clearly better > if you want to do version comparisons. The "extra version" stuff should > be mainly for labeling. The trouble there is that we lack numeric version in some (IMO crucial) places where we only have the string version: - In the startup packet. We send server_version, but not server_version_num, as GUC_REPORT. So if a client wants server_version_num it has to do another round trip to query for it. - In pg_config, where we don't expose any --version-num only --version. -- Craig Ringer http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Craig Ringer <craig@2ndquadrant.com> writes:
> On 12 October 2017 at 06:46, Peter Eisentraut
> <peter.eisentraut@2ndquadrant.com> wrote:
>> I've been using
>> --with-extra-version=+git`date +%Y%m%d`"~"`git rev-parse --short HEAD`
>> for my local builds for some time, and I've not experienced any such
>> problems.
> I've seen issues with a number of tools. The one I can remember most
> clearly is check_postgres.pl . Nobody's going to argue that this is
> pretty code, but last time I tested (9.4-era, admittedly) it exploded
> messily with extra-version.
FWIW, Salesforce tried to do something similar to Peter's example
while I was there. It did not work as well as we'd hoped :-( because
what got baked into the built executables was the latest git commit hash
as of the time you'd last run configure, not what was current as of the
latest "make". Not to mention that you might have built from an
uncommitted state. We tried to find a fix for the former problem that
didn't create lots of overhead, without much success. No idea what
to do about the latter.
These are not big problems for package-building since you basically
never distribute executables that don't get built from a clean state.
But they put a crimp in the usefulness of the idea for developers.
regards, tom lane
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
Hello Tom,
>> I've seen issues with a number of tools. The one I can remember most
>> clearly is check_postgres.pl . Nobody's going to argue that this is
>> pretty code, but last time I tested (9.4-era, admittedly) it exploded
>> messily with extra-version.
>
> FWIW, Salesforce tried to do something similar to Peter's example
> while I was there. It did not work as well as we'd hoped :-( because
> what got baked into the built executables was the latest git commit hash
> as of the time you'd last run configure, not what was current as of the
> latest "make". Not to mention that you might have built from an
> uncommitted state. We tried to find a fix for the former problem that
> didn't create lots of overhead, without much success.
My 0.02€:
For a research project we regenerate a header file with a string
containing the working copy status information,
// file version.h #define REV "<version-output>"
and there is a very small C file which is recompiled with a constant
string based on the version:
// version.c #include "version.h" const char * version = REV;
The make dependencies ensure that the header file is regenerated on each
build with a phony target, and the C file is thus recompiled and linked
into the executables on each build. It means that all executables are
linked on each rebuild, even if not necessary, though.
> No idea what to do about the latter.
"svnversion" adds a "M" for modified on the status. There is an option
with "git describe" to get something similar:
git describe --long --always --all --dirty
Also there is a need of a fall back if this fails, to get "<unknown
version>" instead.
--
Fabien.
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
Hi, On 2017-10-12 22:50:47 +0200, Fabien COELHO wrote: > The make dependencies ensure that the header file is regenerated on each > build with a phony target, and the C file is thus recompiled and linked into > the executables on each build. It means that all executables are linked on > each rebuild, even if not necessary, though. That'd be quite painful, consider e.g. people using LTO. If done right however, that should be avoidable to some degree. When creating the version file, only replace its contents if the contents differ - that's just a few lines of scripting. Greetings, Andres Freund -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
>> The make dependencies ensure that the header file is regenerated on each >> build with a phony target, and the C file is thus recompiled and linked into >> the executables on each build. It means that all executables are linked on >> each rebuild, even if not necessary, though. > > That'd be quite painful, consider e.g. people using LTO. If done right > however, that should be avoidable to some degree. When creating the > version file, only replace its contents if the contents differ - that's > just a few lines of scripting. Indeed. A potential issue is with dynamic linking, potentially someone could recompile/reinstall just one shared object or dll, and the executable using the lib would change its behavior, and run with libs from heterogeneous version. What is the actual version? Hard to say. In dev mode we often use static linking so that we can copy the executable for a previous version and it would not change depending on updated libs, and so that we always know (or should know) what actual version is running. -- Fabien. -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
On Thu, Oct 12, 2017 at 4:50 PM, Fabien COELHO <coelho@cri.ensmp.fr> wrote: > "svnversion" adds a "M" for modified on the status. There is an option with > "git describe" to get something similar: > > git describe --long --always --all --dirty I tried this out a bit: [rhaas pgsql]$ git describe --long --always --all --dirty heads/master-0-gd133982d59 [rhaas pgsql]$ git checkout head~ ...blah blah... [rhaas pgsql]$ git describe --long --always --all --dirty heads/x-218-g6393613b6a Eh, what? Hmm, maybe it's because I have a local branch called 'x' lying around from something or other. [rhaas pgsql]$ git branch -D x Deleted branch x (was 77b6b5e9ce). [rhaas pgsql]$ git describe --long --always --all --dirty tags/REL_10_BETA3-458-g6393613b6a Mmph. I understand the desire to identify the exact commit used for a build somehow, but something whose output depends on whether or not I left a branch lying around locally doesn't seem that great. -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
On Sat, Oct 14, 2017 at 4:47 AM, Robert Haas <robertmhaas@gmail.com> wrote: > Mmph. I understand the desire to identify the exact commit used for a > build somehow, but something whose output depends on whether or not I > left a branch lying around locally doesn't seem that great. Similarly to Peter, I prefer a minimum amount of information so I tend to just use `git rev-parse --short HEAD` with --extra-version for my own builds. Looking at the timestamp of the files installed is enough to know when you worked on them, and when testing a patch and committing it on a local branch before compiling you can know easily where you left things off. git branch --contains is also useful to get from which branch is commit from. -- Michael -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Hello Robert, > Mmph. I understand the desire to identify the exact commit used for a > build somehow, but something whose output depends on whether or not I > left a branch lying around locally doesn't seem that great. Indeed, the branch/tag search might have a little strange behavior. Probably you would be more happy with just the commit (--always) & knowing that it was changed (--dirty). sh> git describe --always --dirty b81eba6 sh> vi README # edit sh> git describe --always --dirty b81eba6-dirty -- Fabien. -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers