Problem creating profiled Root

Dear experts,

I tried to compile Root 5.10.00f with profile support on a SL3 box with gcc 3.2.3. In
config/Makefile.linux I added som '-pg’s, started make and got this error:

utils/src/rootcint_tmp.o(.text+0x542): In function R__tmpnam()': : the use oftmpnam’ is dangerous, better use `mkstemp’
Generating dictionary proofx/src/G__Proofx.cxx…
utils/src/rootcint_tmp -cint -f proofx/src/G__Proofx.cxx -c -Iproofx/inc -I…/xrootd-v20051210-0914/src -Iproofd/inc proofx/inc/TXProofMgr.h proofx/inc/TXProofServ.h proofx/inc/TXSlave.h proofx/inc/TXSocket.h proofx/inc/TXSocketHandler.h
proofx/inc/TXUnixSocket.h proofx/inc/LinkDef.h
gmake: *** [proofx/src/G__Proofx.cxx] Segmentation fault

Does anybody know this error, and is there workaround?

Thanks,
Matthias

Hi,
sounds almost irrelevant, but it works for me on SLC4, GCC 3.4.6, and current CVS. Any one of those might already do it :slight_smile:
Cheers, Axel.

Hi Axel,

but 5.10.00f with gcc 3.2.3 on SL3 is the configuration which we use in BaBar data production.

Matthias

[quote=“Axel”]Hi,
sounds almost irrelevant, but it works for me on SLC4, GCC 3.4.6, and current CVS. Any one of those might already do it :slight_smile:
Cheers, Axel.[/quote]

Hi,
it sounds like a GCC bug that was either fixed later, or is not triggered with later versions of ROOT. As you have to stick to those try removing the -pg from the CINTC(XX)FLAGS, make clean-cint && make.
Cheers, Axel.

Hi Alex,

that was not sufficient, but things work if I “make” the utils before I add the '-pg’s in
Makefile.linux. rootcint_tmp does not work if it was compiled with profile support.

Let’s see now how much time in the BaBar code gets eaten up by Root.

Thanks,
Matthias

[quote=“Axel”]Hi,
it sounds like a GCC bug that was either fixed later, or is not triggered with later versions of ROOT. As you have to stick to those try removing the -pg from the CINTC(XX)FLAGS, make clean-cint && make.
Cheers, Axel.[/quote]

Hi,
why don’t you use kcachegrind? IMHO it’s a lot nicer to use, and it doesn’t require a re-build of the code.
Axel.

Hi Axel,

we started to use cachegrind these days, but people in BaBar want to have gprof as
a second option.

Matthias

[quote=“Axel”]Hi,
why don’t you use kcachegrind? IMHO it’s a lot nicer to use, and it doesn’t require a re-build of the code.
Axel.[/quote]