This is the mail archive of the
mailing list for the Cygwin project.
Re: Mysterious gdb behavior.
- From: "Paul Derbyshire" <derbyshire at globalserve dot net>
- To: cygwin at cygwin dot com
- Date: Sun, 28 Jul 2002 23:48:28 -0400
- Subject: Re: Mysterious gdb behavior.
- References: <3D446034.7038.56AD94A7@localhost>
- Reply-to: derbyshire at globalserve dot net
On 29 Jul 2002 at 12:02, Robert Collins wrote:
> 3) You assert "They do. Windows ALWAYS assigns an 8.3 version of a
> filename that has a long name, a long extension, multiple extensions
> (foo.tar.gz, etc.), or an unusual character in the name such as a
> space." Again, emphasis is mine. We are now both asserting different
And in my experience that's true. If there's a registry hack to make
Windows behave differently, that's neither surprising nor relevant.
Making Windows behave outside of its normal design specifications can
be expected to screw things up. (Hell, even without messing with it
it tends to screw thigns up :)) The default behavior is what's most
important -- anyone who's diddled with it in weird nonstandard ways
ought to be aware it might affect things. Notably any MS-DOS or 16
bit app that needs to access a directory. It's reasonable to expect
it to affect ported stuff in general.
> This disproves your assertions in 1) and 3), and proves 2).
It demonstrates that there's a rare exception that involves purposely
fiddling with the normal behavior of the operating system in a way
that isn't within the default range of behaviors. When someone does
that anything can happen.
> It also demonstrates that you have not followed up the reference
> provided, and are wasting my time.
What reference, provided when? I could hardly follow up that
reference one iteration *before* you posted it. Obviously you again
expected me to read minds.
[Snip patronizing comments about needing to be corrected like a small
I don't care for your tone mister. Also, it's not *my* job to
research all of this stuff. It's my job to use this software and
expect it to work. If something needs to be done to the code to make
it work for a corner case that was not sufficiently allowed for
before, then whoever maintains the code should do it, and whatever
research is needed to make the right changes. I was only suggesting
how it might change. You seem to think I was trying to assert how it
should be changed. I was not. I was speculating. You took my
speculations for facts and then used this straw man to try to make me
look like an idiot on a public forum, and I do not appreciate that. I
suggest you do indeed refrain from answering anything else I post, or
from mentioning me here. Killfiling me might be a good idea.
This whole situation is ludicrous. I posted with a gdb problem.
Someone posted *suspecting* that the directory name with a space in
it was causing the problem. I pointed out that the installer chose
the name, not me, and that I don't recall it giving me a chance to
use another, at least not after informing me that spaces in path
names might cause problems. So I asked what to do about it now. You
suggested editing /etc/passwd but have as yet declined to give me any
more specific instructions, and I'm loath to just assume a simple
thing like changing the name consistently in the file and renaming
the directory will work without a hitch. My other speculations have
proven to have exceptions or other gotchas, as you keep pointing out,
so why should I actually risk reconfiguring my system without enough
knowledge? Aside from your one incomplete (as in, insufficiently
detailed) suggestion you have provided nothing to help my problem,
but you have spewed a great volume of unpleasantness I would have
much rather never heard. In fact, I get the feeling you just like a
good fight, and will try to maneuver people on mailing lists into
having to defend a bogus straw man position you attributed to them
just to amuse yourself. I can think of no other explanation for the
way this has gotten out of hand. One other poster quite helpfully
suggested I just move the project I'm debugging, or even a specific
executable to be debugged, out of the bothersome directory tree; an
excellent idea I wish I'd thought up myself. That poster is a paragon
of helpfulness. You are a paragon of condescending arrogance with no
apparent goal but to subtly reinforce your own belief in your own
intellectual superiority, at the expense of whatever newbie happens
to get in your crosshairs on any given week.
In fact, I think I am about to killfile you. Fire a parting shot if
you wish. I don't care.
> 5) You claim that a Long filename without a short file name will cause
> all sorts of trouble, even though you have been provided with
> documentation from Microsoft about a supported method for creating just
> such a situation.
And it will. The file won't be accessible to legacy applications for
one. And it will cause trouble for cygwin, as *you* pointed out.
There's obviously a very good reason why it's not the default
behavior of the filesystem and why it requires jumping through
registry hoops to do that -- because it occasionally is needed for
something but when messed with casually it screws things up, like
dynamite or highly toxic industrial chemicals.
> You were provided with an external link to authoritative documentation
> about the topic. No guessing was needed about whether SFN are *always*
> created or not.
Before the link was provided, guessing was needed. I had no idea that
link even existed before you posted. What do you expect of me, that I
see the future, or read your mind? Or just that every single thing I
even speculate about I should research for years and get a Ph.D. in
the subject before opening my mouth. (That would, incidentally, cause
a catch-22 when it came time to write the Ph.D. thesis!) The former
is ludicrous because it's obviously beyond the capabilities of most,
if not all, human beings. The latter is ludicrous for essentially the
same reasons. Nobody has time to research every little thing that
pops into their head, or to make every posting they make to a mailing
list a scientific paper with everything properly cited and supported
by cited evidence, graphs, charts, and data points. Even if I had the
Clockstoppers watch and thus had the time I couldn't be arsed to do
so. Besides, my discourse would then make me seem to be Mr. Spock or
something, and I'd quickly know everything, and all conversation
would become boring, and then I'd have no more use for other human
beings, and before long I'd be depressed and lonely and probably kill
Your expectations of a mere user are ridiculous. A user posts saying
they have a problem, they want instructions, as detailed as possible,
on fixes and workarounds, not explanations of how stupid a newbie
they are and why, nor "tough shit, you didn't read the FAQ before you
installed cygwin and its FAQ like you should have", nor a vague "edit
the /etc/passwd file" that would have a genuine stupid newbie type
"nano /etc/passwd" at the bash prompt and then stare in puzzlement at
the screen wondering what to do next.
> In the light of this, I assume that you have not attempted the archival
> search for topic relating to home directories with spaces in them, which
> was the first thing I suggested.
What? I don't remember seeing this suggestion. Besides, it's
ridiculous. I don't need to try it to guess that putting "home
directories with spaces in them" into the search engine will almost
certainly produce exactly one hit: your posting that I'm replying to
now. The odds that that exact phrase was used in the past must be
negligible. Searches merely for "home directories" or for "spaces"
will turn up who knows how many million irrelevant hits. I might as
well read the whole archive from beginning to end. Which, of course,
I refuse to do. It's a lot less work for all concerned if someone who
directory simply answer the question. Putting the fix in the FAQ
even better. Given that Windows has a tendency to generate such
usernames and then Cygwin's insaller uses it by default without
first, lots of people are going to install Cygwin, read the FAQ, and
see that it
says this might cause a problem but not see what to do about it. I
recommend that the FAQ's section on this be expanded to explain how
to change this situation once it's already occurred. (If it hasn't
been already; the version I have might be older.)
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Bug reporting: http://cygwin.com/bugs.html