@EXECSTR line length limit? ... misbehavior?

[vefatica]

Outside TCC, I've been playing with MAPM, the arbitrary precision math library that TCC uses. The command ipow.exe 500 5000 correctly sends the number 500^5000 (13495 characters) to stdout. I thought that would be OK with @EXECSTR but echo %@execstr[ipow.exe 500 5000] produces 13506 characters, and among those characters are 00000000㘴ㄵ〵㠷㘱㌸㔶㘸抵쁎鹷00000000.

Any idea what's going on?

 

[vefatica]

@EXECSTR seems to be able to handle much longer strings, but it produces more of the garbage output. In fact, the garbage output occurs regularly, almost exactly every 8K characters.

 

[rconn]

@EXECSTR has no length limits. It reads the output file in 8K chunks.

 

[vefatica]

@EXECSTR has no length limits. It reads the output file in 8K chunks.

Then it looks like there a problem piecing together buffers.

 

[rconn]

Then it looks like there a problem piecing together buffers.

Not reproducible here. If you have a reproducible failcase, send it & your app to [email protected].

 

[vefatica]

Here's IPOW.EXE (zipped). Run IPOW 500 50000 > file1.txt and echo %@execstr[IPOW.EXE 500 50000] > file2.txt and compare the files. Look near byte 8192. Or just run the second command without redirection and scroll back to see the corruption (might need 2000+ console rows).

v:\> IPOW 500 50000 > file1.txt

v:\> echo %@execstr[IPOW.EXE 500 50000] > file2.txt

v:\> cmp file1.txt file2.txt | head
file1.txt file2.txt differ: char 8193, line 1

 

[vefatica]

Or, if you are wary, do this. (Note: @eval gives the wrong answer)

v:\> echo %@eval[500**50000] > file1

v:\> echo %@execstr[echo %@eval[500**50000]] > file2

v:\> cmp file1 file2
file1 file2 differ: char 8193, line 1

 

[rconn]

The problem here wasn't with @execstr, it was because you are loading a 134K line into a 64K (maximum line size) buffer.

I made a change in 30.0.19 to allow a max line size of 2Gb. But don't do that.

Regarding @eval, it returns the same result as your ipow.exe, provided I use the correct syntax and set the max precision.

 

[vefatica]

The problem here wasn't with @execstr, it was because you are loading a 134K line into a 64K (maximum line size) buffer.

I don't think that's at the heart of the problem. Here's v29 with a rather small file/line.

 

[vefatica]

Regarding @eval, it returns the same result as your ipow.exe, provided I use the correct syntax and set the max precision.

Yeah but that's only because 500^50000 has lots of zeroes at the end. Here's a simple example of @EVAL not doing integer exponentiation as well as it might. This should not end in a 0.

v:\> echo %@eval[2**103]
10141204801825835211973625643010