On 7 August 2011 at 18:54, Kurt Roeckx wrote:
| On Sun, Aug 07, 2011 at 11:10:46AM -0500, Dirk Eddelbuettel wrote:
| >
| > On 7 August 2011 at 17:43, Kurt Roeckx wrote:
| > | On Sun, Aug 07, 2011 at 10:23:09AM -0500, Dirk Eddelbuettel wrote:
| > | >
| > | > On 7 August 2011 at 15:30, Kurt Roeckx wrote:
| > | > | Source: foptions
| > | > | Version: 2140.79-1
| > | > | Severity: serious
| > | [...]
| > | > | xvfb-run: error: Xvfb failed to start
| > | > |
| > | >
| > | > That is a random error in the build infrastructure I can do nothing
about. It
| > | > seems to happen one every couple dozen builds.
| > |
| > | It failed like that one of few one the buildds, that's not one
| > | every dozen.
| > |
| > | So it seems to be random behaviour, but easy to reproduce. In
| > | that case it really shouldn't be that hard to find out what the
| > | difference between working and failing is.
| > |
| > | Could you please try to find out in which package the bug is? I
| > | doubt you're the only one using xvfb to build something, so I
| > | would expect to see more packages failing like that if it was a
| > | bug in xvfb.
| >
| > Kurt, I have been maintaining _several dozen_ of these r-cran-* packages for
| > a half decade or longer. Some of these happen to need an X11 display on
| > startup, so we need xvfb-run.
| >
| > But _not one_ has ever failed xvfb in my pbuilder. So there is no
| > reproduceability at my end.
|
| Ok, that's annoying.
|
| > And if you look at the buildd stats that are logged you see that the
| > autobuilds do succeed most of the time. I really do not know what kills
some
| > of them some of the time. But it is random, and I just don't have any
| > operational hypothesis.
| >
| > I'd love to help, I just have nothing to go by.
| >
| > Any ideas?
|
| Not really.
|
| They now all failed with an other error instead:
| ** preparing package for lazy loading
| Loading required package: MASS
|
| Attaching package: 'fBasics'
|
| The following object(s) are masked from 'package:base':
|
| norm
|
| ** help
| *** installing help indices
| ** building package indices ...
| ** testing if installed package can be loaded
|
| * DONE (fOptions)
| kill: 186: No such process
|
| make: *** [R_any_arch] Error 1
I think the 'kill: 186: No such process' may have been new when me made this
one change in R itself (as the r-cran.mk snippet in every debian/rules is
supplied from the basic R package):
r-base (2.13.0-4) unstable; urgency=low
* debian/r-cran.mk: Make call to xvfb-run more robust by adding '-a'
option to allow automatic switching to free virtual server; with thanks
to Salvatore Bonaccorso for the suggestions (Closes: #630869)
-- Dirk Eddelbuettel <[email protected]> Wed, 22 Jun 2011 08:59:55 -0500
It was failing before too; and other teams seem to be using 'xvfb-run -a' so
I adopted this on Salvatore's suggestion.
I could revert, but maybe we'd just die with a different error message :-/
Dirk
--
Two new Rcpp classes scheduled for New York and San Francisco, details at
http://dirk.eddelbuettel.com/blog/2011/08/04#rcpp_classes_2011-09_and_2011-10
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
