cygstart regression [Was: Re: on the road to 1.5.19 - snapshot testing needed]

Michael Schaap
Mon Oct 3 00:01:00 GMT 2005

On 30-Sep-2005 18:07, Charles Wilson wrote:
> Christopher Faylor wrote:
>> On Fri, Sep 30, 2005 at 09:17:06AM -0600, Aaron Humphrey wrote:
>>> I just did some version testing on the cygstart .mdb problem I 
>>> reported a few
>>> days ago(, and it is
>>> indeed a regression from 1.5.18.  Cygstart works fine with 1.5.18, 
>>> but as
>>> early as the 20050909 snapshot(as far back as I have)it fails on MDB 
>>> files.
>> Sorry but I don't have Microsoft Access so I can't check this.  I 
>> don't consider
>> this a show-stopper either, unfortunately.
>> cygstart comes from cygutils.  I don't remember the genesis of this
>> utility (just because it is in cygutils doesn't mean that Chuck is
>> responsible for it) but it would be nice if the person who supports this
>> utility would comment here.
> cygstart was contributed by Michael Schapp, who is still around but 
> doesn't post often.  cygstart, aside from its option handling, is a 
> very simple app.  The core routine just uses the Windows 
> 'ShellExecute' function on the specified file.  Windows is then 
> responsible for looking up the associated application in the registry, 
> starting it, and causing it to load the specified file.  I don't see 
> how anything in cygwin itself can affect that.
> The ONLY thing I can think of is changes in path handing (conversion 
> between 'unix' and 'windows') as cygstart tries to prepare the target 
> file's pathname for passing into the ShellExecute function -- variable 
> name 'aPath' in the code below.
> But I'll defer to Michael for futher analysis.
As Chuck states, cygstart is a very simple program, and it is indeed 
unlikely that anything in it is causing this problem.

To be able to check whether something has changed in the path handling, 
I've uploaded a version of cygstart that prints out the exact 
ShellExecute call it is making, at 
<>.  Could you perhaps download 
that, try it out both with a snapshot and 1.5.18, and let us know 
whether it prints out something different?

If that's not it, then I don't know what it could be.  Must have 
something to do with timing, or the exact memory location, or something 
like that, especially since the problem disappears under strace...
To track it down, perhaps you could try:
 - "cmd /c start whatever.mdb" from a Cygwin prompt
 - "cygstart whatever.mdb" from a cmd.exe command prompt
 - save the simplest possible Access database, and see if cygstart can 
open that
 - make sure whatever.mdb is on a local disk, in the current directory, 
and has a simple file name

Hope that helps,
  - Michael

Unsubscribe info:
Problem reports:

More information about the Cygwin mailing list