After upgrading to icu-4.4 revdep-rebuild complains about the following lib missing on the system: [ 12% ] * broken /usr/bin/icuinfo (requires libicutest.so.44) It seems like the Makefile{,.in} is missing some lines in the install section. I _blindly_ copy'n'pasted the install section from another Makefile.in and that seems to fix the problem. Reproducible: Always Steps to Reproduce:
Created attachment 224287 [details, diff] icu-4.4-ctestfw-makefilein.patch Please give this file a thorough review. I'm in no way a Makefile-guru and simply copied these lines from another Makefile.in which I found in the icu package.
Created attachment 224289 [details, diff] icu-4.4.ebuild.diff
You are missing inherit autotools there... WFM otherwise. Wondering about the amount of testing this receives upstream before releasing, b/c this is just plain WTF. (And warning for the maintainer, do not attempt full eautoreconf or this piece of junk fails miserably.)
While at it, append-flags -fno-strict-aliasing please, the amount of QA warnings produced is quite alarming.
(In reply to comment #3) > You are missing inherit autotools there... You better have another look at icu-4.4.ebuild.diff again. I didn't forget it.
(In reply to comment #5) > You better have another look at icu-4.4.ebuild.diff again. I didn't forget it. Yeah, I apparently need another bucket of my morning coffee *g*
Someone filed this upstream already (see URL)
*** Bug 310359 has been marked as a duplicate of this bug. ***
Ditto
I suggest to mask =dev-libs/icu-4.4 until it's patched or a new version is pulled in from upstream.
the question is will masking 4.4 cause another problem cuz of deps already rebuilt for this version? - otherwise i'm all for a mask as my webkit-gtk won'T build either..
But webkit-gtk can be installed with icu-4.4 and the Comment webkit-icu-4.4-portage.patch of Comment #18 of Bug #308699.
(In reply to comment #10) > I suggest to mask =dev-libs/icu-4.4 until it's patched or a new version is > pulled in from upstream. Nothing aside from /usr/bin/icuinfo links against libicutest.so.44, as far as I'm aware. So it's annoying, but is it serious enough to warrant a mask? (In reply to comment #11) > the question is will masking 4.4 cause another problem cuz of deps already > rebuilt for this version? > - otherwise i'm all for a mask as my webkit-gtk won'T build either.. Yes, everything you built against the libicu*.so.44 sonames will have to be rebuilt if you go back down to libicu*.so.42. There's a fix for webkit-gtk building with ICU 4.4 in bug #308699.
This bug isn't serious enough to mask =dev-libs/icu-4.4*.
Uh, don't mask this. People will have to recompile lots of compile-intensive stuff yet again (notably, Openoffice).
(In reply to comment #15) > Uh, don't mask this. People will have to recompile lots of compile-intensive > stuff yet again (notably, Openoffice). > Then mask it to save them from compiling openoffice twice (once for the update of icu and once for the downgrade, because icu is (a) broken and (b) because webkit-gtk doesn't work with it, either)!
(In reply to comment #16) > Then mask it to save them from compiling openoffice twice (once for the update > of icu and once for the downgrade, because icu is (a) broken and (b) because > webkit-gtk doesn't work with it, either)! How about that someone goes and commits the darned patches instead? Masking doesn't fix any issue.
i did not have to re-compile openoffice but xulrunner / mozilla stuff and sqlite
(In reply to comment #18) > i did not have to re-compile openoffice but xulrunner / mozilla stuff and > sqlite Well, that's kinda very weird and sounds like OO.org bundling its own icu yet again despite depending on >=dev-libs/icu-4, a.k.a. a bug.
wellto be honest i didn'T try to use OOo but revdep-rebuild dit not find anything to do with it afaik so anyway .. does the patch attached here work now or not? if yes i vote for adding it .. ;)
(In reply to comment #17) > (In reply to comment #16) > > Then mask it to save them from compiling openoffice twice (once for the update > > of icu and once for the downgrade, because icu is (a) broken and (b) because > > webkit-gtk doesn't work with it, either)! > > How about that someone goes and commits the darned patches instead? Masking > doesn't fix any issue. > Please don't get me wrong - that would be my favourite solution as well! I suggested masking icu, because patching seems to take ages :(
The patch works (I'm using it on all of my machines) but as long as there's not even a single comment from upstream about the bug and a proper fix (I still don't know if the patch I attached here is the right solution) we cannot simply add a patch that might lead to other (maybe even more severe) problems. If you cannot wait till this is fixed, just copy the ebuild to your local overlay and apply the patch there.
(In reply to comment #22) > The patch works (I'm using it on all of my machines) but as long as there's not > even a single comment from upstream about the bug and a proper fix (I still > don't know if the patch I attached here is the right solution) we cannot simply > add a patch that might lead to other (maybe even more severe) problems. > If you cannot wait till this is fixed, just copy the ebuild to your local > overlay and apply the patch there. > Good Idea Here works fine thank you!
thank you :)
Seems to be accepted as a bug as of today.
Seems like upstream included the patch: https://bugs.icu-project.org/trac/changeset/27878 But the bug is still open (for review?) so maybe we should still wait with fixing this in our package...
Please add these fixes to portage ASAP. I was having all kinds of problems with this and couldn't figure it out. Posted in the forums and got no help. I even went as far as to rebuild my entire system over the weekend to try and fix revdep-rebuild to no avail. Then I found this bug report. What a waste of 32 hours that I could have avoided if I had found this earlier! If it's already been included upstream then can't we just go ahead and ad it now? It's a heck of a lot simpler to rebuild this than have to rebuild Open Office! On my machine it takes damn near 6 hours to rebuild OO. Again, please fix this ASAP. TIA
patch works fine on AMD64 and i386 platform
*** Bug 312741 has been marked as a duplicate of this bug. ***
The patch also works for me.
Thanks for patch. It works.
patch is working for me.
I think I'm running into this same problem. Every time I revdep-rebuild, it re-emerges dev-libs/icu because /usr/bin/icuinfo requires libicutest.so.44., which is missing. I'm on ~amd64. [ebuild R ] dev-libs/icu-4.4 USE="-debug -doc -examples" 0 kB
Patch fixed the issue I've had for the past weeks, too. That is, revdep-rebuild always rebuilding dev-libs/icu.
Patch works for me as well on ~amd64
arfrever can I get your permission to finally fix this bug? It's been open long enough now and upstream has scheduled the fix for icu-4.4.1 so no need to wait any longer IMHO...
Fixed in dev-libs/icu-4.4-r1.
http://bugs.icu-project.org/trac/ticket/7567 upstream bug has been reviewed and targetted for release in http://bugs.icu-project.org/trac/milestone/4.4.1 Note, this library should not be required for this particular task, and may be removed from install in the future. Mea culpa.
Also, I will try to watch here for downstream bugs. Thanks for filing the bug in ICU. (srl, Tech lead, ICU for C/C++)