Hi Tim and everyone,
I guess I am the person who should be responsible for any anomally in
BlastLib since I was the only person checking in and out from cvs
frequently lately.
I checked out a few copies of blast onto blast02. Did find
BlastLib2/Makefile.in was not up to date, checked in a fresh one. Other
than that, I had no problem compiling the whole blast.
I had a confliction on blast/configure even though I have not change the
file for a while. This means someone probably did not update his cvs
before checking in his change and did not manually resolve the confliction
on his copy of code locally. So some of the changes on configure might
have been reversed. Also, since configure.in did not generate confliction,
I suspect, configure.in and configure are not synchronized. So next time
you do autoconf, the configure file generated would not be the same as the
one in CVS.
I have an explanation for the bloat of code size: if you
look at ~blast/cvs_v2/Makefile.inc, it says CXXFLAGS = -02 -g. I added
option -g since I v been using gdb on BlastLib, the standard is only
-02. I just compiled on blast02 without -g flag, the size is 2268604.
Also, you may find a bloat of size on nsed. If this happened, please check
the BlastLib2/Makefile, if the line reads :
$(bin): %: %.o $(lib) $(objects)
change it back to:
$(bin): %: %.o $(lib)
Again, I did this because in order to use gprof to profile $(bin), all
object files must be linked into it otherwise individual function calls
defined in the object file could not be profiled. I do not know if this is
the right solution or not, but it works.
I remember recompile a libBlast.so on Sept 11, because of the date, and
copied it into ~blast/lib/spud. But I am not sure if the one on Sept 26 was
done by me but it is very possible. the one on Sept 26 has been the one we
used since then.
Cheers
Chi
On Fri, 11 Oct 2002, Timothy Smith wrote:
>
> Hello People,
>
> I am going to try and get the version of BlastLib2
> (version 2) in CVS inline with what is being used and working.
> A few observations. The libBlast.so used by the spuds is
> dated Sept 26, and is 5.4 MegaBytes. I think this is the one
> people are using.
>
> All other versions, includeing the one from cvs
> (cvs up -r v2) at about 2.4-2.6 MegaBytes. Something major
> must be different.
>
> Also I am a bit out of the loop. Clearly there are
> some scripts which are current and critical, and some which
> are dated and my not reflect the important features of libBlast.C .
> So what are they? I am starting by assuming that
>
> /home/blast/blast/commis/phase1/ntuple.C
>
> is still a critical piece.
>
> I think getting cvs right again is something I can do
> from Dartmouth, but I need to know where things stand first.
> If you email me, send you phone number and I'll call you -
> or you can call me.
>
> Tim
> ____________________________________________________________________
> Timothy Paul Smith Assistant Research Professor
> Dartmouth College timothy.p.smith@dartmouth.edu
> Department of Physics and Astronomy -or- tim_smith@mit.edu
> 6128 Wilder Lab.
> Hanover, New Hampshire 03755 tel: (603) 646-9346
>
This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:28 EST