Hacker News new | past | comments | ask | show | jobs | submit login
# - the Unix truth (cwi.nl)
36 points by jluxenberg on July 21, 2010 | hide | past | favorite | 16 comments

'This usually leads to screenfuls of error messages and users praying that in this megabyte of binary garbage no valid shell commands occur' hit a nerve.

Even more terrifying. During my early days of Unix usage, I accidentally (through tab completion and incompetence) executed `source ~/.bash_history`. The outcome was not good.

A similar problem is accidentally dumping binary data to the terminal screen (via an incautious grep, for example), as various bytes get rendered as their corresponding control characters. This is even worse on windows (or at least used to be, I think Win 7 fixed this) where ^G is hard-wired to the system speaker which is fairly difficult to mute effectively.

That's a bit more innocent though, since reset(1) usually helps.

Yeah I had a good laugh because it explains (at least some) of the times I've seen my shell behave like this.

>If there is such nonblank text then for [most unices] this group consists of precisely one argument, as in the example above where argi consists of the single argument "-a -b".

Interesting. This seems pretty unintuitive, I would expect the "-a", "-b" behaviour. (I recognise that this would be more complicated to implement and make it impossible to pass an argument with spaces.) It seems like it could easily cause problems especially if you use `env` to handle paths. But I don't think I've ever run afoul of this.

I miss the old FreeBSD parameter behavior when I could have shebang lines like:

  #!/usr/local/bin/sudo -u someuser /bin/sh

It does work this way, though.

Try the following files:

    #!/usr/bin/sed -f

    #!/usr/bin/sed -e 's/a/b/g' -f

It's not clear to me what point you're trying to make, but the second one gives an error: "-e expression #1, char 2: unknown command: `''", because it's getting a single argument containing a quote.

Removing the quotes gives "-e expression #1, char 10: unknown option to `s'", because "s/a/b/g -f" is not a valid sed command.

And to illustrate my point about env, sed for me is in /bin. But "#!/usr/bin/env sed -f" doesn't work, because it looks for a program named "sed -f".

I was agreeing with you, and providing a simple testcase to show that you were, indeed, right (because I agreed that the behaviour is somewhat surprising). Sorry if that wasn't clear.

> Many Unix systems will consider a `\r' in a DOS-type line-ending (`\r\n') part of the interpreter arguments. (HP-UX, Solaris, UnixWare ignore trailing `\r'. AIX, BSD/OS, FreeBSD, Linux, OpenBSD, Tru64 do not.)

"/bin/sh^M: bad interpreter: No such file or directory" is what Linux gives nowadays after a shell script is edited in Windows' Notepad. I have an idea it used to be more cryptic.

Probably a common failure point, if people use small local scripts in their work-flow, and happen to tweak one in a Windows machine.

There is at last one more case: binfmt_misc on Linux, which gives you Windows-style run-based-on-file-extension.

I had a run-in with it on a shared host with some PHP cron jobs (don't ask). The hosting company changed the handler for *.php and everything broke until I renamed the files. The whole story:


Interesting that he uses a space after the #!. I don't think I've ever done so.

From the article:

    It is rumored that some systems only recognize an 
    executable script when it starts with the four bytes
    `#! /', probably because the GNU autoconf manual says so,
    but it seems impossible to find confirmation for this rumor.

I actually took a weeklong shell course through work where the instructor said not to put whitespace in there.

However, I've never read of any systems that have a problem with whitespace, so I suppose it's just as well to include it.

Yeah, that's what prompted me to even notice, since I was assuming it was just spaced out to make it clearer.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact