Summary: | net-print/hplip-3.16.3: error: relink 'cupsext.la' with the above command before installing it | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | manwe <gentoo> |
Component: | Current packages | Assignee: | Daniel Pielmeier <billie> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | bagzy92, dschridde+gentoobugs, nickAristocrates, printing |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
See Also: | https://launchpad.net/bugs/1589366 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: | log |
Description
manwe
2016-03-22 18:35:21 UTC
Forgot about USE flags. USE="X hpcups libnotify snmp static-ppds -doc -fax -hpijs -kde -libusb0 -minimal -parport -policykit -qt4 -scanner" PYTHON_TARGETS="python2_7" Try if disabling parallel build makes hplip compile: MAKEOPTS="-j1" emerge =net-print/hplip-3.16.3 Yup, works fine. Does compiling on multiple cores fail all the time or just randomly? I've tried to re-emerge it now without -j1 and it worked. So I'd say it's random. As you are the only one who ever reported such an issue and I have never run into this myself I'd say it was only a little hiccup. > I'd say it was only a little hiccup.
OK, let's assume this. Although it happend 3 consecutive times before I filled that bug. I'll reopen this bug if/when needed.
(In reply to Daniel Pielmeier from comment #6) > As you are the only one who ever reported such an issue and I have never run > into this myself I'd say it was only a little hiccup. I run into the same issue (bug #585006). If this is a parallel build failure, it will naturally be very hard to reproduce, but is still very annoying and should be fixed anyway. I now also experienced this issue. However I think it only happened to me if there was a previous failed build of hplip. After cleaning PORTAGE_TMPDIR the build completes without errors. it's no quite understood why cleaning of tmpdir resolves this, because, as known, parallel make failure are hard to reproduce, so it's very easy you are just not facing this in another attempt. Since, this reported more than once, a good workaround for everyone would be restricting makeopts in ebuild during compilation. This is good approach rather closing each report with worksforme. (In reply to Oleg from comment #10) > it's no quite understood why cleaning of tmpdir resolves this, because, as > known, parallel make failure are hard to reproduce, so it's very easy you > are just not facing this in another attempt. Since, this reported more than > once, a good workaround for everyone would be restricting makeopts in ebuild > during compilation. This is good approach rather closing each report with > worksforme. This is just what I observed. Previously I did not have this issue so I was not able to reproduce. Now I am and I found it fails to build repetitive and after cleaning the tmpdir it succeeded all the time. I will eventual add that workaround but you could as well open an upstream bug at hplip. It is possible that this issue only requires small changes in Makefile.am. As you may see I already reopened the bug after it was confirmed by others, so what? so what: emake -j1 install (because it failing in install phase) and forgot about parallel make install failures in future. All happy. and if you able to identify why parallel make installation failing on more low level, report upstream or apply the patch, before this do emake trick so, this is installed for everyone in all cases. Because ebuilds are not designed to fail in random order. *** Bug 585006 has been marked as a duplicate of this bug. *** (In reply to Oleg from comment #12) > so what: > emake -j1 install (because it failing in install phase) and forgot about > parallel make install failures in future. All happy. Yeah this is what will happen the hack is applied and forgotten is about it. emake -j1 install is an ugly workaround and even a QA issue. If I apply this workaround it will fix building it but not the root cause. (In reply to Oleg from comment #13) > and if you able to identify why parallel make installation failing on more > low level, report upstream or apply the patch, before this do emake trick > so, this is installed for everyone in all cases. Because ebuilds are not > designed to fail in random order. I am not able to identify the reason this is just what I found after a short research. Also I am the maintainer of the hplip ebuild and although I could report this issue which is clearly not related to the ebuild itself YOU could do this as well. I am a volunteer and I do not have unlimited time to spend for Gentoo issues. So if you want this bug to get fixed you or anybody else could help and report this on the hplip bug tracker and link it to this issue here. a logic is simple: 1. ebuild must compile and install. 2. if there is upstream troubles in make files, then its a problem indeed, however: 3. you do have power of ebuild system that will fix make troubles. This is a true and respected approach. i find it extremely wrong policy about thinking that emake workarounds considered as hacks or QA issues. Instead of digging for extented time for real cause of parallel make install, you do fix (or rather workaround) on distribution level. Then, you can spend your free time or report upstream in real cause. Thats the simple logic many developers having hard time understanding. bear in mind: distribution ships packages that do install. i find it acceptable to apply any kinds of fixes that following above rule. Suggesting reducing makeopts and in fact, suggesting anything to users to do something by their hands is extremely bad policy. You are not the one who telling users doing things, you are the one who serving users by having more knowledge in package management and distribution internals. So they need their hplip installed in whatever way you fix it. Not suggestions or ugly workarounds that *users* should do. (In reply to Oleg from comment #18) > bear in mind: > distribution ships packages that do install. i find it acceptable to apply > any kinds of fixes that following above rule. Suggesting reducing makeopts > and in fact, suggesting anything to users to do something by their hands is > extremely bad policy. You are not the one who telling users doing things, > you are the one who serving users by having more knowledge in package > management and distribution internals. So they need their hplip installed in > whatever way you fix it. Not suggestions or ugly workarounds that *users* > should do. Man, I already reopened the bug which means I acknowledge that there is an issue. I will deal with it when find the time. Can we please leave it at that. If it is that urgent for you fix it in Funtoo in the meantime. Also ran into this. It would be nice to see a serial build forced in the ebuild, but I also don't have the time to submit a patch at the moment. *** Bug 591428 has been marked as a duplicate of this bug. *** (In reply to Daniel Pielmeier from comment #15) > emake -j1 install is an ugly workaround and even a QA issue. says who? QA project? seems to me that having ebuilds that don't install is a QA issue. devmanual says in multiple places that "emake -j1" is an acceptable workaround for upstream problems, although notes that this will significantly increase compile times for many users. comment and commit. leave bug open if you want, "IN_PROGRESS InVCS". There is a QA check which complains about calling emake with -j1. So it is a QA issue. I already filed a bug about it at the hplip bug tracker to get this fixed properly. I will fix this issue when I am back from holiday. Fixed in GIT. You haven't fixed this. You only made a change in the 3.16.7 ebuild which you later removed. Please fix ALL the ebuilds unless you have some reason to believe this is fixed in the newer versions of hplip. This still happening more than a year later and seriously affects the overall quality of gentoo. I assume that adding an emake -j1 install produces a QA notice because it would be nice if you didn't have to disable parallel make. It is however ENTIRELY unacceptable to be leaving something that just doesn't compile for us. Please stop fighting us and just take the pragmatic fix and disable parallel make for anything not known to be fixed. Parallel build has been fixed for every version starting from 3.16.7! I have not removed it later! It could be applied retroactively for 3.16.3 but I think it would make more sense to stabilise a newer version instead. I am not fighting anybody. I do things I deem right and I already changed my mind here and applied the workaround. The final fix is up to upstream and is tracked there. Nothing here is entirely unacceptable. I think you are making a mountain out of a molehill here. Test the newer versions including the fix and if it works well for you file a stable request bug. Plus the tone in this bug does not very motivate me working for you as a volunteer. Everything we do here we do in our spare time so please keep this in mind. Maybe I made a mistake by fixing this earlier but now it is, so please let go. *** Bug 635490 has been marked as a duplicate of this bug. *** |