Is it normal that two different versions of perl produce different results? - eclipse

Question

I am trying to perform some stack analysis on an MCU following the steps described here. The site links then to a Perl script that I launch as a post-build operation by meanings of a simple batch file.
The IDEA based on Eclipse uses the Perl executable at the path:
C:\..\S32DS_ARM_v2018.R1\utils\msys32\usr\bin\perl.exe
perl.exe -v gives:
This is perl 5, version 22, subversion 1 (v5.22.1) built for i686-msys-thread-multi-64int
The OS (windows) has a perl installation at
C:\Perl64\bin\perl.exe
perl.exe -v gives:
This is perl 5, version 24, subversion 3 (v5.24.3) built for MSWin32-x64-multi-thread
(with 1 registered patch, see perl -V for more detail)
I can confirm that avstak.pl (the perl script I am referring some lines above) produces different results with the former or the latter.
WHY this happens, is out of my area of expertise at the moment.
What I would like to understand is
Understand why this is happening;
Understand which perl provides the right outputs (pretty sure I suppose the 5.24.3 is the correct one);
Learning how to prevent this issue if I am going to use perl in future.
Thanks and best regards,
L.
Edit: the outcome of the script with the two different perl versions (reduced output for readability):
This one is result_5.22.1
Func Cost Frame Height
------------------------------------------------------------------------
> I2C_MasterGetTransferStatus 292 292 1
> FLEXIO_I2C_DRV_MasterStartTransfer 236 236 1
> CLOCK_DRV_Init 172 172 1
> CLOCK_SYS_SetConfiguration 172 172 1
> EDMA_DRV_ConfigScatterGatherTransfer 132 132 1
> CLOCK_SYS_SetSystemClockConfig 76 76 1
> FLEXIO_I2C_DRV_MasterInit 60 60 1
> EDMA_DRV_ConfigSingleBlockTransfer 60 60 1
> main 52 52 1
> LPI2C_DRV_MasterSetBaudRate 52 52 1
> LPI2C_DRV_MasterStartDmaTransfer 52 52 1
> FLEXIO_DRV_InitDriver 52 52 1
> I2C_MasterInit 44 44 1
> LPI2C_DRV_SlaveStartDmaTransfer 44 44 1
> CLOCK_SYS_UpdateConfiguration 44 44 1
> CLOCK_DRV_SetClockSource 44 44 1
> LPI2C_DRV_SlaveInit 44 44 1
> EDMA_DRV_Init 44 44 1
> EDMA_DRV_Deinit 36 36 1
> CLOCK_SYS_ConfigureSOSC 36 36 1
> CLOCK_SYS_ConfigureFIRC 36 36 1
vs
result_5.24.3
Func Cost Frame Height
------------------------------------------------------------------------
> main 536 52 9
I2C_MasterSendDataBlocking 484 28 8
> I2C_MasterReceiveDataBlocking 484 28 8
> I2C_MasterReceiveData 468 20 7
> I2C_MasterSendData 468 20 7
FLEXIO_I2C_DRV_MasterReceiveDataBlocking 456 28 7
FLEXIO_I2C_DRV_MasterSendDataBlocking 456 28 7
FLEXIO_I2C_DRV_MasterSendData 448 20 5
FLEXIO_I2C_DRV_MasterReceiveData 448 20 5
FLEXIO_I2C_DRV_MasterStartTransfer 428 236 4
> I2C_MasterGetTransferStatus 408 292 6
CLOCK_SYS_UpdateConfiguration 336 44 6
CLOCK_SYS_SetConfiguration 292 172 5
> CLOCK_DRV_Init 292 172 5
LPI2C_DRV_MasterReceiveDataBlocking 256 20 7
> I2C_SlaveReceiveDataBlocking 252 12 8
> I2C_SlaveSendDataBlocking 252 12 8
As you can see the hight number in the first version doesn't increase (and it should).
Cost and frame suffer the same issue I suppose.
the script is here:
#!/usr/bin/perl -w
# avstack.pl: AVR stack checker
# Copyright (C) 2013 Daniel Beer <dlbeer#gmail.com>
#
# Permission to use, copy, modify, and/or distribute this software for
# any purpose with or without fee is hereby granted, provided that the
# above copyright notice and this permission notice appear in all
# copies.
#
# THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL
# WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED
# WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE
# AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
# DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
# PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
# TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
# PERFORMANCE OF THIS SOFTWARE.
#
# Usage
# -----
#
# This script requires that you compile your code with -fstack-usage.
# This results in GCC generating a .su file for each .o file. Once you
# have these, do:
#
# ./avstack.pl <object files>
#
# This will disassemble .o files to construct a call graph, and read
# frame size information from .su. The call graph is traced to find, for
# each function:
#
# - Call height: the maximum call height of any callee, plus 1
# (defined to be 1 for any function which has no callees).
#
# - Inherited frame: the maximum *inherited* frame of any callee, plus
# the GCC-calculated frame size of the function in question.
#
# Using these two pieces of information, we calculate a cost (estimated
# peak stack usage) for calling the function. Functions are then listed
# on stdout in decreasing order of cost.
#
# Functions which are recursive are marked with an 'R' to the left of
# them. Their cost is calculated for a single level of recursion.
#
# The peak stack usage of your entire program can usually be estimated
# as the stack cost of "main", plus the maximum stack cost of any
# interrupt handler which might execute.
use strict;
# Configuration: set these as appropriate for your architecture/project.
my $objdump = "arm-none-eabi-objdump";
my $call_cost = 4;
# First, we need to read all object and corresponding .su files. We're
# gathering a mapping of functions to callees and functions to frame
# sizes. We're just parsing at this stage -- callee name resolution
# comes later.
my %frame_size; # "func#file" -> size
my %call_graph; # "func#file" -> {callees}
my %addresses; # "addr#file" -> "func#file"
my %global_name; # "func" -> "func#file"
my %ambiguous; # "func" -> 1
foreach (#ARGV) {
# Disassemble this object file to obtain a callees. Sources in the
# call graph are named "func#file". Targets in the call graph are
# named either "offset#file" or "funcname". We also keep a list of
# the addresses and names of each function we encounter.
my $objfile = $_;
my $source;
open(DISASSEMBLY, "$objdump -dr $objfile|") ||
die "Can't disassemble $objfile";
while (<DISASSEMBLY>) {
chomp;
if (/^([0-9a-fA-F]+) <(.*)>:/) {
my $a = $1;
my $name = $2;
$source = "$name\#$objfile";
$call_graph{$source} = {};
$ambiguous{$name} = 1 if defined($global_name{$name});
$global_name{$name} = "$name\#$objfile";
$a =~ s/^0*//;
$addresses{"$a\#$objfile"} = "$name\#$objfile";
}
if (/: R_[A-Za-z0-9_]+_CALL[ \t]+(.*)/) {
my $t = $1;
if ($t eq ".text") {
$t = "\#$objfile";
} elsif ($t =~ /^\.text\+0x(.*)$/) {
$t = "$1\#$objfile";
}
$call_graph{$source}->{$t} = 1;
}
}
close(DISASSEMBLY);
# Extract frame sizes from the corresponding .su file.
if ($objfile =~ /^(.*).o$/) {
my $sufile = "$1.su";
open(SUFILE, "<$sufile") || die "Can't open $sufile";
while (<SUFILE>) {
$frame_size{"$1\#$objfile"} = $2 + $call_cost
if /^.*:([^\t ]+)[ \t]+([0-9]+)/;
}
close(SUFILE);
}
}
# In this step, we enumerate each list of callees in the call graph and
# try to resolve the symbols. We omit ones we can't resolve, but keep a
# set of them anyway.
my %unresolved;
foreach (keys %call_graph) {
my $from = $_;
my $callees = $call_graph{$from};
my %resolved;
foreach (keys %$callees) {
my $t = $_;
if (defined($addresses{$t})) {
$resolved{$addresses{$t}} = 1;
} elsif (defined($global_name{$t})) {
$resolved{$global_name{$t}} = 1;
warn "Ambiguous resolution: $t" if defined ($ambiguous{$t});
} elsif (defined($call_graph{$t})) {
$resolved{$t} = 1;
} else {
$unresolved{$t} = 1;
}
}
$call_graph{$from} = \%resolved;
}
# Create fake edges and nodes to account for dynamic behaviour.
$call_graph{"INTERRUPT"} = {};
foreach (keys %call_graph) {
$call_graph{"INTERRUPT"}->{$_} = 1 if /^__vector_/;
}
# Trace the call graph and calculate, for each function:
#
# - inherited frames: maximum inherited frame of callees, plus own
# frame size.
# - height: maximum height of callees, plus one.
# - recursion: is the function called recursively (including indirect
# recursion)?
my %has_caller;
my %visited;
my %total_cost;
my %call_depth;
sub trace {
my $f = shift;
if ($visited{$f}) {
$visited{$f} = "R" if $visited{$f} eq "?";
return;
}
$visited{$f} = "?";
my $max_depth = 0;
my $max_frame = 0;
my $targets = $call_graph{$f} || die "Unknown function: $f";
if (defined($targets)) {
foreach (keys %$targets) {
my $t = $_;
$has_caller{$t} = 1;
trace($t);
my $is = $total_cost{$t};
my $d = $call_depth{$t};
$max_frame = $is if $is > $max_frame;
$max_depth = $d if $d > $max_depth;
}
}
$call_depth{$f} = $max_depth + 1;
$total_cost{$f} = $max_frame + ($frame_size{$f} || 0);
$visited{$f} = " " if $visited{$f} eq "?";
}
foreach (keys %call_graph) { trace $_; }
# Now, print results in a nice table.
printf " %-30s %8s %8s %8s\n",
"Func", "Cost", "Frame", "Height";
print "------------------------------------";
print "------------------------------------\n";
my $max_iv = 0;
my $main = 0;
foreach (sort { $total_cost{$b} <=> $total_cost{$a} } keys %visited) {
my $name = $_;
if (/^(.*)#(.*)$/) {
$name = $1 unless $ambiguous{$name};
}
my $tag = $visited{$_};
my $cost = $total_cost{$_};
$name = $_ if $ambiguous{$name};
$tag = ">" unless $has_caller{$_};
if (/^__vector_/) {
$max_iv = $cost if $cost > $max_iv;
} elsif (/^main#/) {
$main = $cost;
}
if ($ambiguous{$name}) { $name = $_; }
printf "%s %-30s %8d %8d %8d\n", $tag, $name, $cost,
$frame_size{$_} || 0, $call_depth{$_};
}
print "\n";
print "Peak execution estimate (main + worst-case IV):\n";
printf " main = %d, worst IV = %d, total = %d\n",
$total_cost{$global_name{"main"}},
$total_cost{"INTERRUPT"},
$total_cost{$global_name{"main"}} + $total_cost{"INTERRUPT"};
print "\n";
print "The following functions were not resolved:\n";
foreach (keys %unresolved) { print " $_\n"; }
Edit2:
As Amon suggested to check, subsequent iterations of the script on the same dataset doesn't produce the same output. Values (cost/frame/height) are always the same but the order in which the functions are reported is different.