XWin.exe crashes

[Harold L Hunt II]

Oh, now this is just great.  I opened this email in Mozilla, hit reply, 
changed the address to send this only to Alan, and it still got sent to 
the mailing list.  Time for a new nightly of Mozilla...

Harold

Harold L Hunt II wrote:

> Alan,
>
> Do you want to apply Alexander's patch directly, or do you want me to 
> apply it locally, test it, then submit it for inclusion in the trunk? 
> Note, this applies to xkb stuff.
>
> Harold
>
> Alexander Gottwald wrote:
>
>> Christopher Faylor wrote:
>>
>>  
>>
>>> I thought your patch only added a "b" to reads and writes which would
>>> indicate that it needed to operate entirely in "binmode".  However, I
>>> see that in some cases, it doesn't add the "b".  Not sure why.
>>>
>>> This seems to be the opposite of what you're implying though.
>>>   
>>
>>
>> I'm not sure for the other cases. So I just changed the case where the
>> "wb" was definitly missing.
>>
>> Fixing bugs at work showed me one thing: Changing not exactly the 
>> condition
>> where the bug occured will likely introduce a new one. If the XFree 
>> maintainer which applies the patch is sure about the other cases, he 
>> can easily add the
>> "t" where it belongs.
>>
>>  
>>
>>> cgf
>>>
>>> Index: programs/Xserver/xkb/ddxLoad.c
>>> ===================================================================
>>> RCS file: /cvs/xc/programs/Xserver/xkb/ddxLoad.c,v
>>> retrieving revision 3.30
>>> diff -u -r3.30 ddxLoad.c
>>> --- programs/Xserver/xkb/ddxLoad.c    2002/05/31 18:46:06    3.30
>>> +++ programs/Xserver/xkb/ddxLoad.c    2002/11/03 12:29:17
>>> @@ -413,7 +413,7 @@
>>>     else if (strlen(xkm_output_dir)+strlen(mapName)+5 <= PATH_MAX)
>>>         sprintf(buf,"%s%s.xkm",xkm_output_dir,mapName);
>>>     if (buf[0] != '\0')
>>> -        file= fopen(buf,"r");
>>> +        file= fopen(buf,"rb");
>>>     else file= NULL;
>>>     }
>>>     else file= NULL;
>>>   
>>
>>
>> This is the server. It reads the binary file. So the "b" should be 
>> there. The automode maybe fixed this automaticly, but this is the 
>> IMHO the better
>> way.
>>
>>
>>  
>>
>>> Index: programs/xkbcomp/xkbcomp.c
>>> ===================================================================
>>> RCS file: /cvs/xc/programs/xkbcomp/xkbcomp.c,v
>>> retrieving revision 3.17
>>> diff -u -r3.17 xkbcomp.c
>>> --- programs/xkbcomp/xkbcomp.c    2002/06/05 00:00:37    3.17
>>> +++ programs/xkbcomp/xkbcomp.c    2002/11/03 14:57:04
>>> @@ -873,16 +873,30 @@
>>>          * -- Branden Robinson
>>>          */
>>>         int outputFileFd;
>>> +        int binMode = 0;
>>> +        const char *openMode = "w";
>>>         unlink(outputFile);
>>> +#ifdef O_BINARY
>>> +        switch (outputFormat) {
>>> +            case WANT_XKM_FILE:
>>> +                binMode = O_BINARY;
>>> +                openMode = "wb";
>>> +                break;
>>> +            default:
>>> +                binMode = 0;
>>> +                break;
>>> +        }
>>> +#endif
>>>         outputFileFd= open(outputFile, O_WRONLY|O_CREAT|O_EXCL,
>>> -                S_IRUSR|S_IWUSR|S_IRGRP|S_IWGRP|S_IROTH|S_IWOTH);
>>> +                
>>> S_IRUSR|S_IWUSR|S_IRGRP|S_IWGRP|S_IROTH|S_IWOTH|binMode);
>>>         if (outputFileFd<0) {
>>>             ERROR1("Cannot open \"%s\" to write keyboard description\n",
>>>                                 outputFile);
>>>             ACTION("Exiting\n");
>>>             exit(1);
>>>         }
>>> -        out= fdopen(outputFileFd, "w");
>>> +        +        out= fdopen(outputFileFd, openMode);
>>>         /* end BR */
>>>         if (out==NULL) {
>>>             ERROR1("Cannot open \"%s\" to write keyboard description\n",
>>>   
>>
>>
>> This is for xkbcomp. Here the binary file gets written. So I added "wb"
>> where it was needed and left the other unchanged.
>> bye
>>    ago
>>  
>>
>