New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Upgrade R to 3.4.0 #23067
Comments
Branch: u/charpent/R340 |
Author: Emmanuel Charpentier |
Commit: |
comment:2
Merged with #22989 (which ==> New commits:
|
comment:3
Note : I am not sure that we are still justified to keep This is unnecessary on Linux. I seem to remember that this was introduced to ease cygwin and/or OS/X compilation. Is that still the case ? |
comment:4
Forgot to add : This R passes its own test suite ( Note that in order to pass it, you HAVE to have en_GB.utf8 configured in your locales. This is also true of the original unpatched R. Note also that the test suite passes much more slowly in Sage than it does on the original unpatched R ; I do not yet know why. |
This comment has been minimized.
This comment has been minimized.
comment:5
Wups ! Forgot to point to the upstream tarball... Fixed now. |
comment:6
Replying to @EmmanuelCharpentier:
I guess so. Did anything change upstream? |
This comment has been minimized.
This comment has been minimized.
comment:7
This installs cleanly on Debian testing and on Ubuntu-over-WSL, but fails to install on cygwin. Reason nunuderstandable (the installation of the recommended packages "stats" seems to need under Cygwin a post-compilation step unneeded on Linux, which fails : rebasing needed ?) ==> Help needed from people knowing cygwin better than I do. Also : no news from the Maoc OS X front (I don't have that in my zoo...), and Apple's shenanigans with SSL have caused trouble for R in the past... |
comment:8
I'm having a look at the Cygwin build issue. |
comment:9
FWIW the failure I get is:
No idea what any of this means. Still investigating. |
comment:10
Replying to @embray:
That's exactly the error I got (see the logs I sent ~ one week ago...).
Neither do I. ISTR that I thought it might be a rebasing problem, but I can't remember how I went to this guess, nor what was its rationale...
Thank you very much : I'm out of my depth, here... |
comment:11
Well for starters there's a bug in R here. The
I don't fully know what that means, but we are in fact building a standard package (the "stats" package) so this file, apparently, isn't guaranteed to exist. The fact that later code uses a variable that isn't guaranteed to be defined is a bug. Why this is being a problem on Cygwin and not on other platforms I don't know. |
comment:12
So on Linux I'm able to reproduce this bug by going into the built source, rm'ing Now just need to figure out where "package.rds" gets generated and see why that's not happening on Cygwin. |
comment:13
Well this escalated quickly. Turns out the "package.rds" files were not being written due to a segfault occurring during their processing, in the Even worse, for some reason R's SIGSEGV handler isn't working and it causes R to exit with a zero return code, indicating success. Weirdly, when I try to find out what was up with that, it seems like something is going horribly wrong in the signal handling. When I try to run it in gdb it seems like R's SIGSEGV handler is not being called at all. But when I run the process through strace, its SIGSEGV handler is called. So there's some shenanigans with the signal stack going on. Finally, this doesn't break R's |
comment:14
Replying to @embray:
Does this happen
Same question : is this Cygwin-specific, or can it be reproduced on Linux (the successful compilation of R being just an happenstance) ?
This is a bug, notwithstanding it gets triggered on Linux or not. Don't you think that it should be reported upstream ? |
comment:15
I agree, the latter issue is a bug, as is the potentially undefined variable I mentioned earlier. The other two issues seem to be Cygwin specific. R is using |
comment:16
Well I can't do anything more with this today. I've traced the segfault to somewhere deep inside the |
comment:17
There's a patch from Cygwin that purports to fix the segfault in pcre. I'm trying it out now. |
comment:18
I should add, the bug causing the segfault makes such a mess of the stack that it probably explains why the segfault wasn't being handled properly either, though I haven't examined the specifics closely, and it's probably not worth the trouble to. |
Dependencies: #23291 |
comment:19
Added #23291, which seems to fix the immediate build issue on Cygwin. Still needs to test that the package actually works. |
comment:20
Well, the R interpreter at least starts up fine and seems to work. That doesn't mean it's free of bugs, but the spkg-check for R is already known to have failures (#22866) so I think any other problems are outside the scope of this ticket, so long as it builds. |
comment:21
I support this statement. |
comment:23
It looks good to me. I'm not sure what you're saying you'll do tonight though. |
Reviewer: Erik Bray |
Changed reviewer from Erik Bray to none |
comment:24
BTW : are you aware of a Mac user who might be solicited to test this on Mac OS X : Apple's shenanigans wit SLL (an, more generally, Xcode) has caused touble in the past... Dima Pasechnik comes to mind, but seems awfukly busy nowadays... Replying to @EmmanuelCharpentier:
|
Reviewer: Erik Bray |
comment:25
Replying to @embray:
Just checking that starting from a fresh VM, I can obtain a working system (and install a few dozen R packages I happen to use more or less constantly). But I agree that's just kicking the tyres... I also think that a similar tyre-kicking might be useful on Macs. NB : the deletion of "Reviewer" has been done by Trac, not by me ! Setting it back. |
Changed branch from u/charpent/R340 to |
Usual reasons (working on an u-to-date R is a prerequisite to get answers from the R Core Team on r-help...)
The new upstream version has been released almost a month ago.
Upstream tarball : https://cran.r-project.org/src/base/R-3/R-3.4.0.tar.gz
Depends on #23291
CC: @embray @jpflori
Component: packages: standard
Keywords: r-project
Author: Emmanuel Charpentier
Branch/Commit:
344d140
Reviewer: Erik Bray
Issue created by migration from https://trac.sagemath.org/ticket/23067
The text was updated successfully, but these errors were encountered: