Created attachment 516512 [details] merge outut failing due to nvidia-drivers blocking other binaries Extremely annoying. I masked the update for nvidia drivers. Something is trigger a rebuild. I am not even using or needing these drivers. I get they may effect X. They should have NO effect on a bunch of binary packages, or other binaries in general.... All the binaries were build on another system. Nothing needs them... Why --keep-going fails makes no sense. Seems like a regress in portage. I have had the drivers not building at various times for months. This is the first time I am able to not merge anything past the failure. I will proceed by manually listing the binaries for update... Dealing with pre-made binaries seem to be extremely troublesome. Why it is so difficult to simply do, emerge -uDNqv world on one system, and then on an identical other system do emerge -uDNkqv world. Always misses stuff, has issues etc. Again all these binaries can merge without needing nvidia-drivers. Why its blocking update makes no sense, it simply should not prevent further packages from merging.
All packages in the list merged fine listing them and doing emerge -qkv1 nvidia-drivers should not have blocked them.
Sorry, I should have added a period. I did not do emerge -qkv1 nvidia-drivers, I did... emerge -qkv1O --jobs 2 --keep-going dev-java/tomcat-servlet-api:4.0 dev-java/tomcat-juli dev-java/tomcat-jni dev-java/jai-core-bin dev-java/jna dev-java/icu4j dev-java/tomcat-jaspic-api dev-java/tomcat-annotations-api dev-java/json-simple dev-java/commons-daemon dev-java/stax2-api dev-java/jctools-core dev-java/conversant-disruptor dev-java/commons-csv dev-java/jcommander dev-java/disruptor dev-java/jnacl dev-java/lz4-java dev-java/lucene-core dev-java/j2objc-annotations dev-java/error-prone-annotations dev-java/httpcomponents-core dev-java/tomcat-util dev-java/tomcat-api dev-java/scribejava-core dev-java/javax-xml-rpc-api dev-java/jackson-annotations dev-java/kafka-clients dev-java/jeromq dev-java/spotbugs-annotations dev-java/tomcat-util-scan dev-java/scribejava-apis dev-java/jackson-databind dev-java/stringtemplate dev-java/tomcat-coyote dev-java/javax-ejb-api dev-java/jackson-dataformat-yaml dev-java/jackson-module-jaxb-annotations dev-java/ant-core dev-java/antlr dev-java/tomcat-catalina dev-java/tomcat-catalina dev-java/jackson-dataformat-xml dev-java/jdbc-jaybird dev-java/log4j-core dev-java/berkeley-db-je dev-java/jgroups dev-java/commons-pool dev-java/commons-dbcp app-crypt/pinentry media-video/rage net-misc/modemmanager media-libs/gegl net-misc/openssh gimp udisks ansible gnucash virtualbox mplayer networkmanager sudo shadow modemmanager gegl consolekit gnupg git ffmpeg xine-lib =x11-libs/wxGTK-3.0.3 =x11-libs/wxGTK-3.0.3-r300 And they all merged. I am pretty sure a regression, as I have never ran into such before.
> x11-base/xorg-drivers:0 > > (x11-base/xorg-drivers-1.19:0/0::gentoo, binary scheduled for merge) conflicts with > x11-base/xorg-drivers[video_cards_nvidia] required by (x11-misc/bumblebee-3.2.1_p20170123:0/0::gentoo, installed) > ^^^^^^^^^^^^^^^^^^ It looks like the above issue is what triggered the nvidia-drivers rebuild. > * <x11-base/xorg-server-1.19.99:0/1.19.5= pulled in by: > * (x11-drivers/nvidia-drivers-387.22:0/387::gentoo, installed) Since the nvidia-drivers rebuild failed, it's understandably upset that the installed nvidia-drivers instance has a broken dependency. > * dev-java/javax-transaction-api:0 pulled in by: > * (dev-java/commons-dbcp-2.2.0:2/2::os-xtoo, binary scheduled for merge) > * > * dev-java/javax-transaction-api:0 pulled in by: > * (dev-java/javax-ejb-api-3.2:0/0::os-xtoo, binary scheduled for merge) These dev-java/javax-transaction-api:0 dependencies are puzzling, since I didn't see dev-java/javax-transaction-api pulled into the merge list. Is that a build-time dependency?
I know what triggered nvidia rebuild. Just not why a rebuild would prevent binaries from merging. dev-java/javax-transaction-api:0 is both build and runtime dep for some binary packages being merged. I am constantly rebuilding lots of Java packages. I do not think that had anything to do with it failing. I guess maybe it could. But other packages did not need the java stuff. I would have thought --keep-going to have worked around the nvidia-drivers, and merged all the rest. I took the same package list, toss it on command line and it merged all binaries fine.
I think I figured out whats going. It seems if you are using binary packages (--usepkg), --keep-going will stop at the first failure, or has some problems when using --usepkg with --keep-going. Merging some java packages that failed, it prevented other unrelated system updates. Doing that same again less binaries, I see it re-calculating deps on the failed packages and dropping others. While the build keeps going.