Buffer Overflow in prs2net for 64-bit Multiplier

[nbingham1]

Buffer overflow when attempting to spice a 64-bit dadda multiplier.

git clone https://github.com/nbingham1/act.git
cd act
git checkout new_act
git submodule update --init --recursive
cd deps
make
cd ..
source setup.py
./build.py clk mult dadda
cd test/prs/clk_mult_dadda___64_64_6______bits/
prs2net -Tst28 -DDUT=true -DLAYOUT=true -DPRSIM=false -B -p "testbench<>" test.act

*** buffer overflow detected ***: prs2net terminated
Aborted (core dumped)

[rmanohar]

This is a known issue (10240 character limit on the output of any identifier component that is being printed out), and is low priority at present.

[rmanohar]

Also, I get the following at the submodule update step in an attempt to replicate.

fatal: clone of 'https://github.com/nbingham1/script.git' into submodule path 'ROOT/act/deps/script' failed

[nbingham1]

Odd, I did recently push to script, maybe you attempted just before the push? I just ran through the steps again and they work on my end :/ Either way, low priority bug...

Perhaps instead of adding all the template parameters to the name, we can just use a unique id after the process name?

[nbingham1]

Here's a test file to use if that continues to be a problem:

test.act

import "qdi/source.act";
import "qdi/sink.act";
import "qdi/fifo.act";
import "bd/fifo.act";
import "clk/fifo.act";
import "gate/delay.act";
import "clk/mult.act";

defproc testbench(globals g, gwrap; bool clk_d, clk; pair A[64]; pair B[64]; pair S[128])
{
    [ DUT ->
    clk::mult_dadda<64, 64, 6> sub;
    g = sub.g;
    sub.clk = clk;
    [ PRSIM ->
    clk = clk_d;
    [] ~PRSIM ->
        ::gate::amplify_fixed<2, NO_INVERT, 6, 384> clk_amp(g, clk_d, clk,);
    ]
    A = sub.A;
    B = sub.B;
    S = sub.S;
    ]
}
globals g, gwrap;
gwrap.sReset = g.sReset;
gwrap._sReset = g._sReset;
gwrap.pReset = g.pReset;
gwrap._pReset = g._pReset;
bool Reset = g.sReset;
bool clk_d, clk;
pair A[64];
pair B[64];
pair S[128];
testbench dut;
g = dut.g;
gwrap = dut.gwrap;
dut.clk_d = clk_d;
clk = dut.clk;
A = dut.A;
B = dut.B;
S = dut.S;

[rmanohar]

Weird. Still having trouble cloning https://github.com/nbingham1/script.git/ Is this private, perhaps?

[rmanohar]

Odd, I did recently push to script, maybe you attempted just before the push? I just ran through the steps again and they work on my end :/ Either way, low priority bug...

Perhaps instead of adding all the template parameters to the name, we can just use a unique id after the process name?

There's a trade-off between small parameter count templated definitions versus ones with a very large number of parameters. We can switch to unique ids if there are more than a certain number of parameters. In any case, this should get fixed once I can replicate the error (the core library should be using snprintf so it shouldn't error out like this in any case, so there must be some location where this was missed).

[rmanohar]

I used the example and found a few sprintfs that should have been snprintfs. At least on my computer it appears to be working now. Pushed a fix.

[nbingham1]

Yep, you fixed it :)