Separate analysis from O2 core?

Dear all,
I have a problem in compiling O2 in a CENTOS7 system. It fails when compiling the newly added analysis tasks, with the error written at the bottom of this mail.
In run2 we decided to split the simulation/reconstruction part (AliRoot), that needed to be stable, from the analysis one (AliPhysics), which undergoes a quick and continuous development.
Wouldn’t it be better to follow the same philosophy also in O2?

Thanks in advance,
best regards,
Diego

P.S. The here is the error message, in case someone can find a quick fix in the meanwhile:
DEBUG:O2:O2:o2-master: [ 96%] Building CXX object Analysis/Tasks/CMakeFiles/O2exe-analysis-jetfinder.dir/jetfinder.cxx.o
DEBUG:O2:O2:o2-master: /home/daq/alice/sw/SOURCES/O2/o2-master/0/Analysis/Tasks/jetfinder.cxx:87:3: internal compiler error: Segmentation fault
DEBUG:O2:O2:o2-master: }
DEBUG:O2:O2:o2-master: ^
DEBUG:O2:O2:o2-master: 0xb4024f crash_signal
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/toplev.c:337
DEBUG:O2:O2:o2-master: 0x7a88d3 lookup_page_table_entry
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/ggc-page.c:630
DEBUG:O2:O2:o2-master: 0x7a88d3 ggc_set_mark(void const*)
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/ggc-page.c:1527
DEBUG:O2:O2:o2-master: 0x700901 gt_ggc_mx_lang_tree_node(void*)
DEBUG:O2:O2:o2-master: ./gt-cp-tree.h:133
DEBUG:O2:O2:o2-master: 0x6efda0 gt_ggc_mx(deferred_access_check&)
DEBUG:O2:O2:o2-master: ./gt-cp-semantics.h:36
DEBUG:O2:O2:o2-master: 0x6efd82 void gt_ggc_mx<deferred_access_check>(vec<deferred_access_check, va_gc, vl_embed>)
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/vec.h:1110
DEBUG:O2:O2:o2-master: 0x6efd82 gt_ggc_mx_vec_deferred_access_check_va_gc_(void
)
DEBUG:O2:O2:o2-master: ./gt-cp-semantics.h:28
DEBUG:O2:O2:o2-master: 0x68d7d7 gt_ggc_mx_tree_check(void*)
DEBUG:O2:O2:o2-master: ./gt-cp-parser.h:38
DEBUG:O2:O2:o2-master: 0x68d842 void gt_ggc_mx<cp_token>(vec<cp_token, va_gc, vl_embed>)
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/vec.h:1110
DEBUG:O2:O2:o2-master: 0x68d842 gt_ggc_mx_vec_cp_token_va_gc_(void
)
DEBUG:O2:O2:o2-master: ./gt-cp-parser.h:49
DEBUG:O2:O2:o2-master: 0x68d8ac gt_ggc_mx_cp_lexer(void*)
DEBUG:O2:O2:o2-master: ./gt-cp-parser.h:76
DEBUG:O2:O2:o2-master: 0x68da0d gt_ggc_mx_cp_parser(void*)
DEBUG:O2:O2:o2-master: ./gt-cp-parser.h:136
DEBUG:O2:O2:o2-master: 0x916435 ggc_mark_root_tab
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/ggc-common.c:77
DEBUG:O2:O2:o2-master: 0x916710 ggc_mark_roots()
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/ggc-common.c:94
DEBUG:O2:O2:o2-master: 0x7a92c7 ggc_collect()
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/ggc-page.c:2197
DEBUG:O2:O2:o2-master: 0x80d0e4 cgraph_node::finalize_function(tree_node*, bool)
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/cgraphunit.c:480
DEBUG:O2:O2:o2-master: 0x6ed6df expand_or_defer_fn(tree_node*)
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/cp/semantics.c:4279
DEBUG:O2:O2:o2-master: 0x6af8c4 cp_parser_function_definition_after_declarator
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/cp/parser.c:26293
DEBUG:O2:O2:o2-master: 0x6afba4 cp_parser_late_parsing_for_member
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/cp/parser.c:27166
DEBUG:O2:O2:o2-master: 0x690ed5 cp_parser_class_specifier_1
DEBUG:O2:O2:o2-master: …/…/gcc/gcc/cp/parser.c:22347
DEBUG:O2:O2:o2-master: Please submit a full bug report,
DEBUG:O2:O2:o2-master: with preprocessed source if appropriate.
DEBUG:O2:O2:o2-master: Please include the complete backtrace with any bug report.
DEBUG:O2:O2:o2-master: See https://gcc.gnu.org/bugs/ for instructions.
DEBUG:O2:O2:o2-master: gmake[2]: *** [Analysis/Tasks/CMakeFiles/O2exe-analysis-jetfinder.dir/build.make:63: Analysis/Tasks/CMakeFiles/O2exe-analysis-jetfinder.dir/jetfinder.cxx.o] Error 1
DEBUG:O2:O2:o2-master: gmake[1]: *** [CMakeFiles/Makefile2:23435: Analysis/Tasks/CMakeFiles/O2exe-analysis-jetfinder.dir/all] Error 2
DEBUG:O2:O2:o2-master: gmake: *** [Makefile:163: all] Error 2

Interesting topic. @ihrivnac This could be a topic for WP3.

@dstocco is this using alidock?

@eulisse: no, I’m compiling with alibuild:

aliBuild -z -j16 --defaults o2 build O2

How many GB of RAM do you have? try -j <number-of-GB>.

ram is not an issue. I’m using an (old) FLP machine with 32 cores and 64 GB of ram.
Limiting myself to 16 only to avoid conflicting with other users (but there is nobody else consuming memory now…)

I am afraid @lmiglior had the same issue yesterday on a machine with plenty of RAM. She solved by updating alidist and O2…

Yes exactly! The installation crashed at 99% and after the updating of alidist and O2, it finished!

Hi @lmiglior and @pezzi,
unfortunately, both my alidist and my O2 are updated to this morning. alidist did not change since then (not sure about O2),
Anyways, for the moment I solved the issue by avoiding compiling Analysis/Tasks…
Thanks!
Cheers,
Diego

GCC from the system?

Nope, apparently it downloads its own:

`/home/daq/alice/sw/slc7_x86-64/GCC-Toolchain/v7.3.0-alice1-13/bin/gcc`

despite I’m explicitly loading devtool 7:

source scl_source enable devtoolset-7