|
What is the complete definition of the dir_struct structure? In Linux the information provided by the readdir(3) - Linux manual page[^] system call provides a properly null terminated string.
|
|
|
|
|
So in my case (in Windows) I am not getting that null on ending string (\0) ?
|
|
|
|
|
How do you obtain the file name from Windows? Which API function call are you using? If the documentation for that API function call specifies that the name is null terminated, then you should not be getting the garbage. But some old file systems set off a fixed length buffer for the file name - even in Unix! The original Unix directory was 16 bytes per entry: two bytes inode, 14 chars filename. The orginal FAT had room for 8 char file name, 3 char extension. If the field was filled with a maximum length file name (in either Unix or FAT), there was no room for a terminating nul. If you reach the end of the field before finding a null, it is nevertheless the end of the field!
Remember that nul terminated strings is an idea conceived for the C language. It directoy conflicts with the ISO character set standard, whether ISO 646, 8859 or other variants. Other languages have used other solutions, e.g. in Pascal, a string is (at least in most implementations) represented by a length field associated with the string buffer (so the ISO NUL character can be used in the ISO standard way!). Many other languages to the same.
DOS was developed before C and Unix. Windows was developed before Unix had become very important, at least in the market segments Windows was aimed at. So you should not take for granted that all sorts of C / Unix conventions have been adopeted in every other segment.
|
|
|
|
|
Obviously not. But you need to look into the original source code to see why.
|
|
|
|
|
#define DIR_NAME_LEN 2048
struct dir_struct
{
void* display;
char directory[DIR_NAME_LEN];
...
....
I have followed the way of how this variable is handled:
dir_struct_t dir_struct;
TRACE(_T("Directory - %s\n"), dir_struct.directory); -> _CrtDbgReport: String too long or IO Error
strncpy(dir_struct->directory, "/", sizeof(dir_struct->directory));
TRACE(_T("Directory - %s\n"), dir_struct->directory); -> /
strcat(dir_struct->directory, current_file->name);
TRACE(_T("Directory - %s\n"), dir_struct->directory); -> _CrtDbgReport: String too long or IO Error
And if I look into VS debugger when I got "_CrtDbgReport: String too long or IO Error" from TRACE macro, I see the following result: see image.[^] and Untitled2 — Postimage.org[^]
The line strcat(dir_struct->directory, current_file->name); is not quite correct ? I mean, correct for Windows, even if is correct for Linux ...
|
|
|
|
|
You are making a wrong assumption. The sizeof(dir_struct->directory) operator will return 2048. But what if the string in that table is less than that? You need an actual count of the number of valid characters in the directory field (assuming that it is correctly null terminated).
You are also mixing the dot (.) and arrow (->) operators on your struct reference.
But all this is largely irrelevant. You need to go back a stage to the code that fills the directory structure in the first place, as that is where the problem is most likely to be found.
|
|
|
|
|
"You are also mixing the dot (.) and arrow (->) operators on your struct reference."
Yes, you are right, I taken those lines of codes from different function (in one of them was reference, in other was pointer), that why those mixing are there.
modified 29-Apr-20 15:24pm.
|
|
|
|
|
Well you should not be doing so, since the reference in both cases is a structure, not a pointer. The dot operator is the correct one.
But more importantly you have still not shown or explained when and where this structure gets initialised, and the directory name copied in. Until you give us that information anything we pass on is pure guesswork.
|
|
|
|
|
You have actually identified your problem, you just haven't realized it.
strncpy(dir_strut->directory, "/", sizeof(dir_struct_directory)); strcat(dir_struct->directory, current_file->name);
It would seem that current_file->name is not a null terminated C string. You have not given us the details of current_file, or how it is instantiated, so we can't further diagnose the problem at the moment. Perhaps current_file has a member that tells you how long the name member is, in which case you should use that and strncat to append the file name to the directory name.
Keep Calm and Carry On
|
|
|
|
|
Yes, indeed, from this point strcat(dir_struct->current_directory, current_file->name);, my dir_struct.directory seem to go crazy.
current_file->name is coming from:
const file_info* current_file = AnotherFunction();
and file_info is a struct defined like this:
typedef struct
{
char* name;
.....
unsigned int status;
}file_info;
Here is the steps:
strncpy(dir_struct->directory, "/", sizeof(dir_struct->directory));
TRACE(_T("Directory - %s\n"), dir_struct->directory); -> Directory - /
Untitled — Postimage.org[^]
const file_info* current_file = AnotherFunction();
TRACE("\n===>>>%s\n", current_file->name); ->===>>>Unbroken
Untitled2 — Postimage.org[^]
TRACE(_T("Directory - %s\n"), dir_struct->directory); -> _CrtDbgReport: String too long or IO Error
Untitled — Postimage.org[^]
Untitled — Postimage.org[^]
|
|
|
|
|
The AnotherFunction itself or the way you are using it seems to be a reason of your problem... 
|
|
|
|
|
Yes Victor, you had right, from AnotherFunction is coming sometime a value like this: Oe©*Bº cE«Mòצ£š{@Q—Y|†Š˜ b!
And this AnotherFunction is list_entry_const() ... is coming from Linux, and to reproduced in a test project is quite difficult ...
Now I have to see where is the starting point for that strange chars ...
modified 3-May-20 10:01am.
|
|
|
|
|
It looks like this function returns some local variable that goes out of scope after returning. However, better would be looking in its sorce code.
|
|
|
|
|
I have just tried to reproduce what you have done and the code works fine. I can only conclude that the code you are showing does not match what you are actually running. But since you have only shown small pieces and not the complete method where this occurs, it is difficult to say more.
|
|
|
|
|
I really, really would show the entire code because I am struggling of it, but unfortunately I cannot due to confidentiality issue. I'll try to write more code here.
|
|
|
|
|
What you need to display is the assignment to dir_struct.directory - nothing more. It may be as simple as a single statement (/line). I guess that confidentiality issues won't preclude that.
How did you get hold of that value? Which system call or library function provided it?
If you say "None, we built it from pieces ourselves, and that code is confidential", then you have got the answer: In that string building code you forgot to add the terminating nul. Otherwise: Tell us who provided the directory string, in which way.
|
|
|
|
|
No, this code is not written by us. I'll try to write here as much as I can.
P.S. The buggy code is written in C for Linux, and I don't know C (I guess this is visible  )
P.S. I don't pretend to know C++ either ... only that I fight in this domain ....
modified 30-Apr-20 12:13pm.
|
|
|
|
|
I am sure that you can modify any confidential pieces so they do not reveal anything that they should not.
|
|
|
|
|
info* current = list_entry(walker, const info, list);
TRACE(_T("current->name %s\n"), current->name); -> _CrtDbgReport: String too long or IO Error
Why I got _CrtDbgReport: String too long or IO Error I wrote in my previous posts ...
modified 30-Apr-20 12:23pm.
|
|
|
|
|
I have managed to create a sample of all that code, although I find the #define list_entry_const somewhat difficult to deconstruct. However, with a valid name in the file_info entry it works fine. I can only assume that somewhere in the actual code that you are running there is some corruption of your data, or a pointer is not set up correctly.
[edit]
I also could not find the TRACE macro in Windows.
[/edit]
modified 30-Apr-20 9:46am.
|
|
|
|
|
Richard MacCutchan wrote:
I also could not find the TRACE macro in Windows.
Isn't that an MFC thing?
"One man's wage rise is another man's price increase." - Harold Wilson
"Fireproof doesn't mean the fire will never come. It means when the fire comes that you will be able to withstand it." - Michael Simmons
"You can easily judge the character of a man by how he treats those who can do nothing for him." - James D. Miles
|
|
|
|
|
Could be, but I haven't used MFC since about 2003, and I don't intend doing so ever again.
|
|
|
|
|
Richard MacCutchan wrote: I also could not find the TRACE macro in Windows.
After a lot of calls... it finishes calling _CrtDbgReportW
|
|
|
|
|
It's an MFC* macro (as pointed out by David Crow above).
*Yes, I know, that combination of letters should no more pass my lips.
|
|
|
|
|
Like they say in the movies, the names have been changed to protect the innocent.
"One man's wage rise is another man's price increase." - Harold Wilson
"Fireproof doesn't mean the fire will never come. It means when the fire comes that you will be able to withstand it." - Michael Simmons
"You can easily judge the character of a man by how he treats those who can do nothing for him." - James D. Miles
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
