IPAddress.Parse("4294967296") does not throw FormatException on OS X

[moander]

System.Net.IPAddress.Parse("4294967296") does not throw FormatException on OS X.

$ uname -a
Darwin 15.6.0 Darwin Kernel Version 15.6.0: Mon Aug 29 20:21:34 PDT 2016; root:xnu-3248.60.11~1/RELEASE_X86_64 x86_64 i386 MacBookPro11,5 Darwin

$ dotnet --version
1.0.0-preview2-003121

Steps to reproduce

mkdir foo
cd foo
dotnet new
dotnet restore
sed -i '9i Console.WriteLine(System.Net.IPAddress.Parse("4294967296"));' Program.cs
dotnet run

The sed command writes Console.WriteLine(System.Net.IPAddress.Parse("4294967296")); to line 9 in Program.cs

IPAddress.Parse returns 0.0.0.0 on my Mac when using dotnet run.

$ dotnet run
Project bug (.NETCoreApp,Version=v1.0) was previously compiled. Skipping compilation.
0.0.0.0
Hello World!

On Windows, Ubuntu, Docker on Mac, and Mono on Mac I get the expected FormatException:

$ mono bin/Debug/netcoreapp1.0/bug.dll 

Unhandled Exception:
System.FormatException: An invalid IP address was specified.

[moander]

Travis did not get an exception either:
https://travis-ci.org/aspnet/BasicMiddleware/jobs/160775559

info:       Expected: False
info:       Actual:   True
fail:     Microsoft.AspNetCore.HttpOverrides.IPNetworkTest.ParseAndTryParse_ShouldNotParse(networkString: "4294967296") [FAIL]
info:       Stack Trace:
info:         /Users/travis/build/aspnet/BasicMiddleware/test/Microsoft.AspNetCore.HttpOverrides.Tests/IPNetworkTest.cs(94,0): at 

[moander]

Ping? :)

[karelz]

@Priya91 can you please triage it offline? What is the impact on 2.0?

[stephentoub]

https://github.com/dotnet/corefx/issues/8361

[karelz]

Dupe of #17219

[stephentoub]

@karelz, there's some new info in this issue though: apparently mono is generating a FormatException... at the very least, I suggest we understand why, e.g. is mono doing its own parsing? is it using a different OS API? etc.

[karelz]

Fair point, I missed that part.

[Priya91]

This is a platform difference on OS X, inet_aton doesn't return invalid address for single part ipaddress that's greater than 32 bits. Instead it returns address of 0.0.0.0. We can identify this behavior and handle the difference in the wrapper.

Also note the Linux documentation states less than 32 bit number is valid for single part, whereas OS X documentation for the API doesn't explicitly state it.

#include <stdio.h>
#include <sys/types.h>
#include <netinet/in.h>
#include <arpa/inet.h>

int main()
{
	char *ipaddress = "4294967296";
	struct in_addr inaddr;

	int result = inet_aton(ipaddress, &inaddr);

	if (result == 0)
		printf("Address is invalid\n");

	else printf("address is %s\n", inet_ntoa(inaddr));

	return 0;
}

return,

lasekar-osx:myprograms lakshmipriya$ ./a.out
address is 0.0.0.0

[karelz]

If it is documented behavior and if it is easy to do in Mac PAL, let's do it.

[akoeplinger]

@stephentoub

e.g. is mono doing its own parsing?

We're actually using referencesource in this case: IPAddress.cs#L307

[stephentoub]

Thanks, @akoeplinger. @CIPop, @pgavlin, why did we move away from this implementation for corefx? i.e. why didn't we keep using the managed implementation that desktop is using?
https://referencesource.microsoft.com/#System/net/System/Net/IPAddress.cs,272

[CIPop]

why did we move away from this implementation for corefx?

I have changed the code in the (very) early componentization stages (before .Net Core) to be compliant with OneCore.
For reasons having to do with RFC compliance, support, code unification and binary size the networking architect as well as the Winsock team recommended the new APIs (e.g. RtlIpv4AddressToStringExW).