[Rd] Compiling R on Windows XP - Rgui crashes yet Rterm works

[Adler, Avraham]

I think I found the issue (through blind luck). When I recompiled without using -march=corei7 and instead leaving it as -mtune=core2 allows the gui to load properly. Now it seems to pass the tools library tests as well.

Thank you.

--Avraham Adler



-----Original Message-----
From: Adler, Avraham 
Sent: Wednesday, May 09, 2012 7:25 PM
To: 'r-devel at r-project.org'
Subject: Compiling R on Windows XP - Rgui crashes yet Rterm works

Hello.

I am trying to build R on Windows. It appears that my build passes the various "make checks" (unless I missed some error) and running Rterm seems to work fine whereas Rgui has an immediate error stating "

	AppName: rgui.exe AppVer: 2.150.58871.0 ModName: rzlib.dll

	ModVer: 0.0.0.0 Offset: 0000a9e5

The taskbar still works, but the console window has nothing in it, and trying to open a script drops me back to Windows.

Even though it seemingly passing make check-all, I ran 'testInstalledBasic ("both")' in the terminal and received the same error that was addressed here in 2010 <https://stat.ethz.ch/pipermail/r-help/2010-May/237947.html> and was not addressed here in 2011 <https://stat.ethz.ch/pipermail/r-help/2011-June/282354.html>: that primitives fail on >=.

I'm not certain if this affects anything, as in Rterm, the >= comparison works fine (4 >= 5 returns FALSE and 4 >= 1 returns TRUE).

Be that as it may, it would be very helpful to find out what would cause the compiled R to have Rgui not work properly yet pass all the checks? I'd rather use the gui than the terminal, for what it is worth.

If it helps, I have been using the following:

Tarball - R-2.15.0.tar.gz
Rtools - 2.15
	CYGWIN 1.7 already installed (for ATLAS) so Rtools CYGWIN dlls NOT installed
	Tcl/tk, libjpeg, libpng, libtiff all installed cairo-current-win.tar.gz

The only other change I made was to set EOPTS=-march=corei7 as the GCC included in the toolchain in 4.6.3 which recognizes the i7 (as opposed to Cygwin which is using 4.5.3 and only recognizes the core2).

My computer is a Lenovo W520 with a Core i7 2760QM running Windows XP Professional.

If there is any other information that would be helpful in uncovering why this is happening, please let me know.


Thank you very much,

Avraham Adler

PS: Copying the compiled Rblas.dll over to my regular R install (binary from the CRAN) seems to work just fine and provides a significant speedup in matrix operations in the few tests that I have done.