During the reconstruction of the gbrahms cdat files in the the MDC-I Alv
found several files that would not work be read properly. The following
is a summary of all files processed
taken from the summary. This by it self is interesting
GEANT Input Events Real CPU Output Comments
Job size proc'd time time size
138 543283320 2492 7:58:33 28646.82 rectype -490668032
139 543686656 2504 6:14:40 22425.16 rectype 877693696
140 540213248 491 segmentation
violation
141 544362496 1668 segmentation
violation
142 518387064 2388 5:21:38 19259.20 rectype 1052452287
143 530743296 2441 3:53:25 13956.22 rectype 0
144 404851676 MIDS
145 404430316 MIDS
146 404566668 MIDS
147 404990604 MIDS
148 228028588 MIDS
149 227641836 MIDS
150 668454772 466 segmentation
violation
151 669371988 74 segmentation
violation
152 1344524 6 0:00:12 10.43
153 668659268 1434 segmentation
violation
154 668444348 1660 3:47:04 13531.55 rectype 0
157 665620212 1751 3:44:41 13435.35 rectype 0
159 666202076 299 segmentation
violation
160 680880132 3 0:00:09 5.03 rectype 0
161 683008204 3 0:02:42 161.58 rectype 0
162 678697820 74 segmentation
violation
163 679257044 671 3:02:04 10895.24 rectype 0
164 81156036 2940 2:16:15 8156.06
165 81289780 2940 1:22:39 4952.06
166 296982732 4900 1:29:08 5328.36
167 302970604 5000 4:52:40 17542.89
168 302978700 835 segmentation
violation
169 303143644 5000 1:41:49 6089.18
The files in question here are those which terminated with an rectype
xxxxx error.
One of the files were investigated in detail namely sim_160.cdat which
dies quickly.
It turns out that the .cdat file is corrupted (for unknown reason) . By
scanning the file longer it in facts turn out that after a while the
data are recognizably again.
The BrGeantInput class was modified so that it skips until a potential
end of
stream (i.e. end of event) Since this marked is a -1 that will also
appear for tracks with q=-1 this search is not unique. If the reading
checks the retry flag by testing on
readflags like the code snippet
if(ginp->TryRepair()){
delete gevent;
continue;
}
he reading seem to contine properly.
The corrections which concerns
BrGeantInput.h and BrGeantInput.cxx has been checked into cvs
repository.
-- ------------------------------------------------------------------Flemming Videbaek Physics department, Brookhaven national laboratory Upton L.I., NY11973, USA
Phone: (516)-344-4106 FAX: (516)-344-1334 secr: (516)-344-7681 e-mail: videbaek@bnl.gov
This archive was generated by hypermail 2b29 : Tue Feb 01 2000 - 20:35:04 EST