Hello Masa,
When I compiled the code using g++ on the linux8664gcc platform using ACLIC fstream::tellg() functions as expected, so I would assume that the compiler knows how to do the pos_type to long int conversion properly.
For now I have made a workaround by compiling the snipit of code that I need to return the correct value from tellg(). This works fine so I don't think it is a C++ limitation, although I understand that for large files it may not work properly since long int cannot handle those numbers.
Thanks
Kerry
>-----Original Message-----
>From: Masaharu Goto [mailto:gotom_at_hanno.jp]
>Sent: Wednesday, May 18, 2005 12:35 AM
>To: LEE, KERRY T. (JSC-SF) (UHCL)
>Cc: ''''roottalk_at_pcroot.cern.ch' ' ' '; '''Philippe Canal ' ' '; 'Fons
>Rademakers '
>Subject: Re: [ROOT] fstream on linuxx8664gcc platform
>
>Hello Kerry,
>
>After a while, I come to think that such usage of tellg() and
>pos_type may not be allowed in C++. It happened to work
>on ICC but the code is not portable. Since file size may be
>beyond the size of long int. And pos_type may contain larger
>number than long int can handle. Hence, conversion from
>pos_type to long int is not defined in C++. The number you
>get from tellg() in g++ is an address of pos_type object.
>
>This is my conclusion, but tell me if I am wrong. If there is
>a standardized conversion from pos_type to long int, I can
>add that interface to gcc3strm.cxx.
>
>Thank you
>Masa Goto
>
>
>
>On 2005/05/18, at 10:25, LEE, KERRY T. (JSC-SF) (UHCL) wrote:
>
>>
>> I found out that if I cut out this piece of code that reads the file
>> and
>> make it a seperate function, then compile it with ACLIC, fstream works
>> fine.
>> This is not so elegant code but it works. I will be looking forward
>> to a
>> more permanent fix from Masa and Philippe.
>>
>> Thanks
>> Kerry
>>
>> -----Original Message-----
>> From: Fons Rademakers
>> To: LEE, KERRY T. (JSC-SF) (UHCL)
>> Cc: '''roottalk_at_pcroot.cern.ch' ' '; ''Masaharu Goto ' '; ''Philippe
>> Canal '
>> '
>> Sent: 5/17/2005 4:48 PM
>> Subject: Re: [ROOT] fstream on linuxx8664gcc platform
>>
>> There is not directly a workaround for this. I hope Masa and Philippe
>> will
>> look into this (and the other stream problem).
>>
>> Cheers, Fons.
>>
>>
>>
>> LEE, KERRY T. (JSC-SF) (UHCL) wrote:
>>> Hi Fons,
>>>
>>> Thanks for this workaround.
>>>
>>> There is another place in my code where I do the following:
>>>
>>> fstream fff("mntExHdd",ios_base::in|ios_base_out);
>>> ....
>>> Do some reading of the file
>>> ...
>>> fff.tellg();
>>>
>>> This last line reports a wrong number just like the size was
>> incorrect. Is
>>> there also a portable way in ROOT to work around this problem?
>>>
>>> Thanks
>>> Kerry
>>>
>>>
>>> -----Original Message-----
>>> From: Fons Rademakers
>>> To: LEE, KERRY T. (JSC-SF) (UHCL)
>>> Cc: ''roottalk_at_pcroot.cern.ch' '; 'Masaharu Goto '; 'Philippe Canal '
>>> Sent: 5/14/2005 2:30 AM
>>> Subject: Re: [ROOT] fstream on linuxx8664gcc platform
>>>
>>> Hi Kerry,
>>>
>>> this will give you file sizes in a portable way in ROOT:
>>> #include <TSystem.h>
>>> FileStat_t s;
>>> gSystem->GetPathInfo("mntExHdd", s)
>>> s.fSize // contains the size
>>>
>>>
>>> -- Fons
>>>
>>>
>>> LEE, KERRY T. (JSC-SF) (UHCL) wrote:
>>>
>>>>
>>>> Hi Fons,
>>>>
>>>> In the meantime, do you know of a workaround I can use to get the
>>>> file
>>>
>>> size?
>>>
>>>> Thanks
>>>> Kerry
>>>>
>>>> -----Original Message-----
>>>> From: Fons Rademakers
>>>> To: LEE, KERRY T. (JSC-SF) (UHCL)
>>>> Cc: 'roottalk_at_pcroot.cern.ch'; Masaharu Goto; Philippe Canal
>>>> Sent: 5/13/2005 7:29 PM
>>>> Subject: Re: [ROOT] fstream on linuxx8664gcc platform
>>>>
>>>> Hi Kerry,
>>>>
>>>> I see the same problem on my AMD64. Looks like a CINT issue. I'll
>>>> forward
>>>> it to Masa.
>>>>
>>>> Cheers, Fons.
>>>>
>>>>
>>>>
>>>> LEE, KERRY T. (JSC-SF) (UHCL) wrote:
>>>>
>>>>
>>>>> Dear Root Team,
>>>>>
>>>>> I have transferred some code from a linuxicc platform ROOT version
>>>>
>>>> 4.02/00
>>>>
>>>>
>>>>> to linuxx8664gcc ROOT version 4.04/02 and it crashed due to
>>>>> incorrect
>>>>> determination of the filesize using fstream. Here is a sample code
>>>>
>>>> that
>>>>
>>>>
>>>>> reproduces the problem, where you just put a file in ascii format in
>>>>
>>>> place
>>>>
>>>>
>>>>> of "mntExHdd". The correct filesize in my case is 105bytes.
>>>>>
>>>>> //TestStream.C
>>>>> {
>>>>> TString FileName="mntExHdd";
>>>>> fstream fff(FileName,ios_base::in);
>>>>> if(fff.good()){
>>>>> fff.seekg(0,ios_base::end);
>>>>> int FileSize = fff.tellg();
>>>>> cout<<"FileSize="<<FileSize<<endl;
>>>>> fff.seekg(0,ios_base::beg);
>>>>> }
>>>>> fff.close();
>>>>> }
>>>>>
>>>>> The results from linuxicc
>>>>>
>>>>> *******************************************
>>>>> * *
>>>>> * W E L C O M E to R O O T *
>>>>> * *
>>>>> * Version 4.02/00 17 December 2004 *
>>>>> * *
>>>>> * You are welcome to visit our Web site *
>>>>> * http://root.cern.ch *
>>>>> * *
>>>>> *******************************************
>>>>>
>>>>> FreeType Engine v2.1.3 used to render TrueType fonts.
>>>>> Compiled for linuxicc with thread support.
>>>>>
>>>>> CINT/ROOT C/C++ Interpreter version 5.15.159, Nov 14 2004
>>>>> Type ? for help. Commands must be C++ statements.
>>>>> Enclose multiple statements between { }.
>>>>> root [0] .x TestStream.C
>>>>> FileSize=105
>>>>>
>>>>>
>>>>> The Results from linuxx8664gcc
>>>>>
>>>>> *******************************************
>>>>> * *
>>>>> * W E L C O M E to R O O T *
>>>>> * *
>>>>> * Version 4.04/02 9 May 2005 *
>>>>> * *
>>>>> * You are welcome to visit our Web site *
>>>>> * http://root.cern.ch *
>>>>> * *
>>>>> *******************************************
>>>>>
>>>>> Compiled for linuxx8664gcc with thread support.
>>>>>
>>>>> CINT/ROOT C/C++ Interpreter version 5.15.169, Mar 14 2005
>>>>> Type ? for help. Commands must be C++ statements.
>>>>> Enclose multiple statements between { }.
>>>>> root [0] .x TestStream.C
>>>>> FileSize=7470896
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> Thanks
>>>>> Kerry
>>>>
>>>>
>>>
>>
>> --
>> Org: CERN, European Laboratory for Particle Physics.
>> Mail: 1211 Geneve 23, Switzerland
>> E-Mail: Fons.Rademakers_at_cern.ch Phone: +41 22 7679248
>> WWW: http://www.rademakers.org/fons/ Fax: +41 22 7679480
>>
Received on Wed May 18 2005 - 17:06:59 MEST
This archive was generated by hypermail 2.2.0 : Tue Jan 02 2007 - 14:45:08 MET