Windows: using hg executable from an IDE results in python errors

[Alexander Belchenko]

Steve Borho пишет:
> On Thu, 2008-03-20 at 11:20 -0700, Howard Ha wrote:
>> Hi everyone,
>>
>> I'm completely new to Mercurial, but a longtime cvsnt user considering 
>> moving our entire team to Hg.
>>
>> We do web development on Windows boxes using PHPEd as our IDE, and 
>> luckily that tool has the ability to integrate with external apps by 
>> executing "shell scripts" (they use this to integrate svn and cvs with 
>> the IDE).  The added benefit of using this over just using tortoiseHg is 
>> that we can assign keyboard shortcuts (highly useful) to the actions 
>> that we integrate (push, pull, update, commit, log, etc).
>>
>> I managed to figure out how to create an appropriate shell script 
>> command line for using Mercurial with PHPEd, but the result is a crash 
>> and error dump.  The EXACT SAME command in a windows command prompt 
>> works perfectly fine.
>>
>> The error output is:
>>
>> c:\development\tortoisehg\hg.exe log benchmark_library.inc.php --cwd 
>> E:\projectscvs\beeza\includes\
>> Traceback (most recent call last):   
>> File "hg", line 29, in <module> 
>> File "mercurial\util.pyc", line 992, in set_binary
>> IOError: [Errno 9] Bad file descriptor
>>
>> I'm using the tortoisehg's hg.exe executable, but the same errors 
>> (different line numbers) come up if I use the executables made by Lee 
>> Cantey (http://mercurial.berkwood.com/).
>>
>> Opening up cmd.exe and running:
>>
>> c:\development\tortoisehg\hg.exe log benchmark_library.inc.php --cwd 
>> E:\projectscvs\beeza\includes\
>>
>> Works perfectly.
>>
>> Any ideas what the problem might be and any possible work arounds?  I'm 
>> unsure how PHPEd is executing the script, but I assume there must be 
>> some environment variables missing or something.
>>
>> Any help would be GREATLY appreciated.
> 
> One of the first thing Mercurial does when invoked on Windows is to set
> the binary attribute on STDIN, STDOUT, and STDERR file handles.  This is
> required in order for commands that input/output text to work correctly
> (import, export, diff, etc).
> 
>    msvcrt.setmode(fd.fileno(), os.O_BINARY)
> 
> The file handles provided by PHPEd do not care for this function call at
> all.  I suspect we should be catching the IOError and ignoring it.
> 
> What do you think, Patrick?
> 
I'm not Patrick but you need to test is fd hasattr fileno. When program running as
subprocess then pipes fd never had fileno. Because they're not real files.