bug-bash
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Bash-4.3-rc1 - problems and other problems


From: Linda Walsh
Subject: Re: Bash-4.3-rc1 - problems and other problems
Date: Sat, 25 Jan 2014 17:36:23 -0800
User-agent: Thunderbird



Chet Ramey wrote:
On 1/24/14, 4:16 PM, Linda Walsh wrote:
Trying to build,

Get undefined reference to `sh_xfree' In function `write_helpfiles'.

Anyone run into this... was it a mistake to try for separate help
files?

Thanks for the report.  It's a simple patch, so I've attached it.
----
FWIW -- my workaround was to remove the "--with-separate-helpfiles" option.

As it seemed to be called in something to do with the helpfiles.






Also of note, a parallel make often fails ...

Should I assume that the bash-make isn't multi-process safe? Or
is that unintentional?

It's just a missing dependency.  I wonder why I haven't run into this
before; I run `gmake -j 4' to build bash on every system I use.
----


I often will use "make -j" for most things....

A limit sometimes will improve efficiency on large builds (ex. kernel),
but difference is negligible w/many builds.


Also, haven't begin to look at "why", but something is alot slower
in bash-43...

Noticed it during login.

Added a -x at beginning of profile.sh with a function
to print "SECONDS+a millisecond counter".

To get to beginning of calling my first "personal file"
(~/.bashrc): bash43 shows:
[ 108.973]/etc/profile#356> test -r /home/law/.bashrc
[ 109.999]/etc/profile#356> . /home/law/.bashrc
-------------
bash42 shows:
[  24.084]/etc/profile#357> test -r /home/law/.bashrc
[  24.090]/etc/profile#357> . /home/law/.bashrc
....finishes in a bit over 31 seconds...

----

Note--- real login time <1 second max in 4.2 -- including
my personal scripts, the 24 & 31 are w/the tracing overhead, so
divide by ~20 -- bash43 seemed like several seconds -- most of which
was in the suse setup.  but haven't narrowed it down anymore
than coming up with the raw data...which I could dump on you
if you wanted it -- up to the point where it starts to call
my .bashrc -- i.e. before that would be all suse's intro scripts.

I've no idea why there is such a large discrepancy at this point.

My build options:
prefix=/
. ~/.config.include
gen_config_ops

declare -a bash_enables=( alias arith-for-command array-variables
  brace-expansion casemod-attributes casemod-expansions command-timing
  cond-command cond-regexp coprocesses debugger directory-stack
  direxpand-default disabled-builtins dparen-arithmetic extended-glob
  extended-glob-default glob-asciiranges-default help-builtin
  history job-control multibyte net-redirections
  process-substitution progcomp prompt-string-decoding readline
  restricted select single-help-strings )

# declare -a bash_with=( curses included-ltdl pic gnu-malloc 
ltdl-lib=/usr/lib64)

  declare -a bash_with=( bash-malloc curses gnu-ld )

export LDFLAGS="-flto -fpie "
for ena in "address@hidden" ; do
  config_ops+="--enable-$ena "
done

for withs in "address@hidden" ; do
  config_ops+="--with-$withs "
done

------------------

config.include sets the directories and CFLAGS w/CFLAGS set to:

declare -a CFLAGS_AR=(
-g -ggdb -O2 -m64 -fasynchronous-unwind-tables -fbranch-target-load-optimize -fdelete-null-pointer-checks -fgcse-after-reload -fgcse-las -fgcse-sm -fgraphite-identity -fipa-pta -fivopts -floop-block -floop-flatten -floop-interchange -floop-strip-mine -flto -fmessage-length=0 -fpredictive-commoning -frename-registers -freorder-blocks-and-partition -ftracer -ftree-loop-linear -ftree-loop-distribution -ftree-loop-distribute-patterns -ftree-loop-im -ftree-loop-ivcanon -ftree-vectorize -ftree-slp-vectorize -funswitch-loops -funwind-tables -fvariable-expansion-in-unroller -fvect-cost-model -fweb -march=native -pipe
        )


(basically all the optimizations from -O3 that don't increase space).

-------
When I noticed the time problem, I made sure bash was
stripped (strip /bin/bash) & made sure it was "prelinked"











reply via email to

[Prev in Thread] Current Thread [Next in Thread]