hydra/doc/dev-notes.txt

* Recreating the schema bindings:

  $ make -C src/sql update-dbix

* Running the test server:

  $ DBIC_TRACE=1 ./script/hydra_server.pl

* Setting the maximum number of concurrent builds per system type:

  $ sqlite3 hydra.sqlite "insert into SystemTypes(system, maxConcurrent) values('i686-linux', 3);"

* Creating a user:

  $ sqlite3 hydra.sqlite "insert into Users(userName, emailAddress, password) values('root', 'e.dolstra@tudelft.nl', '$(echo -n foobar | sha1sum | cut -c1-40)');"

  (Replace "foobar" with the desired password.)

  To make the user an admin:

  $ sqlite3 hydra.sqlite "insert into UserRoles(userName, role) values('root', 'admin');"

  To enable a non-admin user to create projects:
  
  $ sqlite3 hydra.sqlite "insert into UserRoles(userName, role) values('alice', 'create-projects');"
  
* Creating a release set:

  insert into ReleaseSets(project, name) values('patchelf', 'unstable');
  insert into ReleaseSetJobs(isPrimary, project, release, job, attrs, description) values(1, 'patchelf', 'unstable', 'tarball', 'officialRelease=false', 'Source distribution');
  insert into ReleaseSetJobs(project, release, job, attrs, description) values('patchelf', 'unstable', 'build', 'system=i686-linux', 'Build on i686-linux');
  insert into ReleaseSetJobs(project, release, job, attrs, description) values('patchelf', 'unstable', 'build', 'system=x86_64-linux', 'Build on x86_64-linux');
  insert into ReleaseSetJobs(project, release, job, attrs, description, mayFail) values('patchelf', 'unstable', 'rpm_fedora9i386', '', 'Fedora 9 (i386)', 1);
  insert into ReleaseSetJobs(project, release, job, attrs, description) values('patchelf', 'unstable', 'rpm_fedora10i386', '', 'Fedora 10 (i386)');
  insert into ReleaseSetJobs(project, release, job, attrs, description) values('patchelf', 'unstable', 'deb_ubuntu804i386', '', 'Ubuntu 8.04 (i386)');

  insert into ReleaseSets(project, name) values('patchelf', 'stable');
  insert into ReleaseSetJobs(isPrimary, project, release, job, attrs, description) values(1, 'patchelf', 'stable', 'tarball', 'officialRelease=true', 'Source distribution');
  insert into ReleaseSetJobs(project, release, job, attrs, description) values('patchelf', 'stable', 'build', 'system=i686-linux', 'Build on i686-linux');
  insert into ReleaseSetJobs(project, release, job, attrs, description) values('patchelf', 'stable', 'build', 'system=x86_64-linux', 'Build on x86_64-linux');
  insert into ReleaseSetJobs(project, release, job, attrs, description, mayFail) values('patchelf', 'stable', 'rpm_fedora9i386', '', 'Fedora 9 (i386)', 1);
  insert into ReleaseSetJobs(project, release, job, attrs, description) values('patchelf', 'stable', 'rpm_fedora10i386', '', 'Fedora 10 (i386)');
  insert into ReleaseSetJobs(project, release, job, attrs, description) values('patchelf', 'stable', 'deb_ubuntu804i386', '', 'Ubuntu 8.04 (i386)');

* Changing the priority of a scheduled build:

  update buildschedulinginfo set priority = 200 where id = <ID>;

* Changing the priority of all builds for a jobset:

  update buildschedulinginfo set priority = 20 where id in (select id from builds where finished = 0 and project = 'nixpkgs' and jobset = 'trunk');
  

* Steps to install:

  - Install the Hydra closure.

  - Set HYDRA_DATA to /somewhere.
  
  - Run hydra_init.pl

  - Start hydra_server

  - Visit http://localhost:3000/

  - Create a user (see above)

  - Create a project, jobset etc.

  - Start hydra_evaluator and hydra_queue_runner


* Upgrade notes:

  alter table builds add column longDescription text;
  alter table builds add column license text;
  alter table projects add column homepage text;
  alter table builds add column homepage text;
  alter table BuildProducts add column defaultPath text;
  alter table BuildResultInfo add column failedDepBuild integer;
  alter table BuildResultInfo add column failedDepStepNr integer;
  alter table ReleaseSetJobs add column jobset text not null default "trunk";
  === (DB dump/load needed after Sqlite upgrade) ===
  insert into jobs(project, jobset, name, active) select distinct project, jobset, job, 0 from builds b where not exists (select 1 from jobs where project = b.project and jobset = b.jobset and name = b.job);

  create index IndexBuildInputsByBuild on BuildInputs(build);
  create index IndexBuildInputsByDependency on BuildInputs(dependency);
  
  create index IndexBuildsByTimestamp on Builds(timestamp);

  alter table jobs add column disabled integer not null default 0;
  alter table builds add column maintainers text;

  # Add the isCurrent column to Builds and use the obsolete Jobs.active to fill it in.
  alter table builds add column isCurrent integer default 0;
  update builds set isCurrent = 1 where id in (select max(id) from builds natural join (select distinct b.project, b.jobset, b.job, b.system from builds b join (select project, jobset, name from jobs where active = 1) j on b.project = j.project and b.jobset = j.jobset and b.job = j.name) b2 group by project, jobset, job, system);

  alter table Jobsets add column enabled integer not null default 1;

  # Releases -> Views.
  alter table ReleaseSets rename to Views;
  alter table ReleaseSetJobs rename to ViewJobs;
  alter table ViewJobs rename column release_ to view_;
  alter table ViewJobs drop column mayFail;
  alter table ViewJobs add column autorelease integer not null default 0;

  alter table Builds add column nixExprInput text;
  alter table Builds add column nixExprPath text;

  # Adding JobsetEvals.
  drop table JobsetInputHashes;
  (add JobsetEvals, JobsetEvalMembers)

  
* Job selection:
  
    php-sat:build [system = "i686-linux"]
    php-sat:build [same system]
    tarball [same patchelfSrc]
    --if system i686-linux --arg build {...}


* Restarting a bunch of failed builds:

  $ sqlite3 hydra.sqlite "select x.id from builds x join buildresultinfo r on r.id = x.id where project = 'nixpkgs' and jobset = 'stdenv' and exists (select 1 from buildinputs where build = x.id and revision = 14806) and finished = 1 and buildstatus = 3" > ids

  $ for i in $(cat ids); do echo $i; sqlite3 hydra.sqlite "begin transaction; insert into buildschedulinginfo (id, priority, busy, locker) values($i, 100, 0, ''); delete from buildresultinfo where id = $i; update builds set finished = 0 where id = $i; commit transaction;"; done

  Or with Postgres:

  (restarting all aborted builds with ID > 42000)
  $ psql -h buildfarm.st.ewi.tudelft.nl -U hydra hydra -t -c 'select x.id from builds x join buildresultinfo r on r.id = x.id where finished = 1 and buildstatus = 3 and x.id > 42000' > ids

  $ for i in $(cat ids); do echo $i; PGPASSWORD=... psql -h buildfarm.st.ewi.tudelft.nl -U hydra hydra -t -c "begin transaction; insert into buildschedulinginfo (id, priority, busy, locker) values($i, 100, 0, ''); delete from buildresultinfo where id = $i; update builds set finished = 0 where id = $i; commit transaction;"; done
  

* select * from (select project, jobset, job, system, max(timestamp) timestamp from builds where finished = 1 group by project, jobset, job, system) x join builds y on x.timestamp = y.timestamp and x.project = y.project and x.jobset = y.jobset and x.job = y.job and x.system = y.system;

  select * from (select project, jobset, job, system, max(timestamp) timestamp from builds where finished = 1 group by project, jobset, job, system) natural join builds;


* Delete all scheduled builds that are not already building:

  delete from builds where finished = 0 and not exists (select 1 from buildschedulinginfo s where s.id = builds.id and busy = 1);


* Installing deps.nix in a profile for testing:

  $ nix-env -p $NIX_USER_PROFILE_DIR/hydra-deps -f deps.nix -i \* --arg pkgs 'import /etc/nixos/nixpkgs {}'
  

* select x.project, x.jobset, x.job, x.system, x.id, x.timestamp, r.buildstatus, b.id, b.timestamp
  from (select project, jobset, job, system, max(id) as id from Builds where finished = 1 group by project, jobset, job, system) as a_
  natural join Builds x
  natural join BuildResultInfo r
  left join Builds b on b.id =
    (select max(id) from builds c
     natural join buildresultinfo r2
     where x.project = c.project and x.jobset = c.jobset and x.job = c.job and x.system = c.system
           and x.id > c.id and r.buildstatus != r2.buildstatus);

* Using PostgreSQL:

  $ HYDRA_DBI="dbi:Pg:dbname=hydra;" hydra_server.pl


* Find the builds with the highest number of build steps:

  select id, (select count(*) from buildsteps where build = x.id) as n from builds x order by n desc;


* Evaluating the NixOS Hydra jobs:

  $ ./hydra_eval_jobs ~/Dev/nixos-wc/release.nix --arg nixpkgs '{outPath = /home/eelco/Dev/nixpkgs-wc;}' --arg nixosSrc '{outPath = /home/eelco/Dev/nixos-wc; rev = 1234;}' --arg services '{outhPath = /home/eelco/services-wc;}' --argstr system i686-linux --argstr system x86_64-linux --arg officialRelease false


* Show all the failing jobs/systems in the nixpkgs:stdenv jobset that
  succeed in the nixpkgs:trunk jobset:

  select job, system from builds b natural join buildresultinfo where project = 'nixpkgs' and jobset = 'stdenv' and iscurrent = 1 and finished = 1 and buildstatus != 0 and exists (select 1 from builds natural join buildresultinfo where project = 'nixpkgs' and jobset = 'trunk' and job = b.job and system = b.system and iscurrent = 1 and finished = 1 and buildstatus = 0) order by job, system;