It seems the 'all' packages are built for i386 as for amd64. As I understood, they should not be built for i386 and only for amd64. They are built for both architectures.
I'm havling a look in the sql configuration and from the following table, it seems that the configuration is correct: build_architecture_all only for amd64 and i386 is optional (so insert if i386 build fails). 355 CREATE TABLE IF NOT EXISTS "mini_buildd_architectureoption" ("id" integer NOT NULL PRIMARY KEY AUTOINCREMENT, "extra_options" text NOT NULL, "pickled_data" text NOT NULL, "architecture_id" varchar(50) NOT NU LL REFERENCES "mini_buildd_architecture" ("name") DEFERRABLE INITIALLY DEFERRED, "distribution_id" integer NOT NULL REFERENCES "mini_buildd_distribution" ("id") DEFERRABLE INITIALLY DEFERRED, "optional" bool NOT NULL, "build_architecture_all" bool NOT NULL);^M 356 INSERT INTO mini_buildd_architectureoption VALUES(1,'','','amd64',1,0,1);^M 357 INSERT INTO mini_buildd_architectureoption VALUES(3,'','','amd64',2,0,1);^M 358 INSERT INTO mini_buildd_architectureoption VALUES(5,'','','amd64',3,0,1);^M 359 INSERT INTO mini_buildd_architectureoption VALUES(11,'','','amd64',6,0,1);^M 360 INSERT INTO mini_buildd_architectureoption VALUES(13,'','','i386',6,1,0);^M 361 INSERT INTO mini_buildd_architectureoption VALUES(14,'','','i386',3,1,0);^M 362 INSERT INTO mini_buildd_architectureoption VALUES(15,'','','i386',1,1,0);^M 363 INSERT INTO mini_buildd_architectureoption VALUES(16,'','','i386',2,1,0);^M On Thu, 12 Mar 2020 at 11:15, Marc Leeman <marc.lee...@gmail.com> wrote: > This is the status of the openjdk-8 build: as the log confirms, the i386 > packages got inserted, while the amd64 not due to the md5 conflict since > both are creating the same openjdk-8-source_8u242-b04-2~company10+6_all.deb > > In my setup, the error seems consistent for "any" source packages that > also produce an "all" package. From a reprepro point of view, it makes > sense; though I doubt that I am the first one that requires to compile > these packages. I must be missing something. > > mini-buildd@debian-builder01:~/repositories/televic$ reprepro listmatched > buster-televic-unstable '*openjdk*' > buster-company-unstable|main|amd64: openjdk-8-dbg 8u232-b09-1~deb9u1 > buster-company-unstable|main|amd64: openjdk-8-demo 8u232-b09-1~deb9u1 > buster-company-unstable|main|amd64: openjdk-8-doc 8u242-b04-2~company10+6 > buster-company-unstable|main|amd64: openjdk-8-jdk 8u232-b09-1~deb9u1 > buster-company-unstable|main|amd64: openjdk-8-jdk-headless > 8u232-b09-1~deb9u1 > buster-company-unstable|main|amd64: openjdk-8-jre 8u232-b09-1~deb9u1 > buster-company-unstable|main|amd64: openjdk-8-jre-headless > 8u232-b09-1~deb9u1 > buster-company-unstable|main|amd64: openjdk-8-jre-zero 8u232-b09-1~deb9u1 > buster-company-unstable|main|amd64: openjdk-8-source > 8u242-b04-2~company10+6 > buster-company-unstable|main|i386: openjdk-8-dbg 8u242-b04-2~company10+6 > buster-company-unstable|main|i386: openjdk-8-demo 8u242-b04-2~company10+6 > buster-company-unstable|main|i386: openjdk-8-doc 8u242-b04-2~company10+6 > buster-company-unstable|main|i386: openjdk-8-jdk 8u242-b04-2~company10+6 > buster-company-unstable|main|i386: openjdk-8-jdk-headless > 8u242-b04-2~company10+6 > buster-company-unstable|main|i386: openjdk-8-jre 8u242-b04-2~company10+6 > buster-company-unstable|main|i386: openjdk-8-jre-headless > 8u242-b04-2~company10+6 > buster-company-unstable|main|i386: openjdk-8-jre-zero > 8u242-b04-2~company10+6 > buster-company-unstable|main|i386: openjdk-8-source 8u242-b04-2~company10+6 > buster-company-unstable|main|source: openjdk-8 8u242-b04-2~company10+6 > > On Thu, 12 Mar 2020 at 11:04, Marc Leeman <marc.lee...@gmail.com> wrote: > >> >> As it turns out, I have had the issue before and did not really notice it. >> >> I tried to backport openjdk8 to buster and both builds (after fixing >> lintian) errors succeed. However, there is that strange reprepro error as >> described before (so not from a portext this time). >> >> The result is that the i386 package gets inserted (not mandatory) but the >> amd64 package does not end up in the repository. The log entries are again >> (see log), but it is again the md5 conflict that is acting up. >> >> >> >> >> On Tue, 25 Feb 2020 at 13:36, Stephan Sürken <abs...@debian.org> wrote: >> >>> Hi Marc, >>> >>> On Fri, 2020-02-21 at 11:54 +0100, Marc Leeman wrote: >>> > >>> > Could it be that there is some config option that is still wreaking >>> > havoc after having been disabled. >>> >>> Fwiw, there is this 'inconvenience' bug: >>> >>> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=838393 >>> >>> So just be sure to restart the Daemon after configuration changes. >>> >>> > The initial setup might have had amd64/i386 build architecture all >>> > enabled. >>> > >>> > I then disabled i386 to get something working quickly only to enable >>> > it again a couple of months later, this time with the build >>> > architecture all disabled for i386. >>> >>> Hmm, ok. Seems suspicious, but would still not explain the behaviour >>> you are seeing, imho. >>> >>> Could you, before doing the failing portext, check the repo status for >>> the package? I.e., s.th. like >>> >>> --- >>> su - mini-buildd >>> cd ~/repositories/<repo>/ >>> reprepro listmatched stretch-<repo>-experimental '*gstreamer*' >>> --- >>> >>> Thx! >>> >>> S >>> >>> >> >> -- >> g. Marc >> > > > -- > g. Marc > -- g. Marc