Platform: Windows

[Kelmar]

Platform: Windows
Arch: x64
Compiler: VS2015

When running node with --eval or --require without the required argument,
node would segfault. This should correct that and display a reasonable
error message again.

[thefourtheye]

@Kelmar I couldn't reproduce the problem

➜  node git:(master) node -e 'console.log(process.version)'
v6.2.0
➜  node git:(master) node --eval
node: --eval requires an argument
➜  node git:(master) node --require
node: --require requires an argument
➜  node git:(master) node --evail
node: bad option: --evail

Edit: Do you mean, this is specific to Windows?

[addaleax]

It looks to me like this is a bug, but it might be better to fix it in node_main.cc: In this line, there is no additional slot for the trailing nullptr allocated, although it probably should be. The fact that argv[argc] == NULL is pretty much something one should be able to take for granted.

[Kelmar]

@addalex That is probably a more correct fix yes. The assumption seems to be that argv would be terminated with a nullptr, but apparently wasn't the case when I mistakenly did a "node -e" command and got a segfault.

[addaleax]

@Kelmar Would you be interested in updating this PR with that fix or creating a new one?

[thefourtheye]

argv is not a single string of characters, but an array of character pointers. argc specifies the length of the array. Why would additional slot for nullptr be needed?

[addaleax]

@thefourtheye Yeah, I know. It’s standard-mandated that argv terminates with nullptr (see e.g. http://stackoverflow.com/a/16419021/688904 )… it’s definitely not needed, but it’s just something that coders may expect to hold true.

[Kelmar]

@addaleax I would be happy to make the change. :)

[addaleax]

@Kelmar Great! Let us know if you need anything. :)

[thefourtheye]

Ah, okay. If the standard mandates then lets do it. cc @bnoordhuis

[Kelmar]

I have created a new pull request arg_segfault_fix_2 :)

[addaleax]

Closing this in favour of #6938 then :)