追加输出文件的时候奇怪的错误 [英] Strange bug when appending output to file

问题描述

我有追加在bash以文件的形式获取文本即使我告诉它删除追加时，奇怪的错误。

我有一个bash脚本，我做以下

 回声运行程序＆GT; foo.txt的
./fortran_program＆GT;＆GT; foo.txt的
 

该计划fortran_program（它的一个精简版，让相同的结果）是：

 程序的主
写一个'
写（*，*）'B'
程序结束
 

这应执行后给我运行程序AB 在foo.txt的。不过我觉得操作B 代替原文已被覆盖和 A 不包括在内。

我做了 strace的-f ./script ，发现下面的输出，看起来相关的：

  ...
[PID 36681]打开（foo.txt的，O_WRONLY | O_CREAT | O_APPEND，0666）= 3
[PID 36681] dup2（3,1）= 1
[PID 36681]关闭（3）= 0
...
[PID 34260]写（1，A \\ N，3）= 3
[PID 34260] lseek的（1，0，SEEK_CUR）= 3
[PID 34260] ftruncate（1,3）= 0
[PID 34260]写（1，B \\ N，3）= 3
...
 

从我十分有限的这种理解似乎 A 写入文件，然后 lseek的被称为这给该文件只有3个字符（长度相同的第一个字符串我尝试写），然后 ftruncate 截断它在3个字符（即运行），然后写入 B 该文件。

在C以下程序（回声运行程序＆GT; foo.txt的; ./c_program ）抄录所以现在看来​​，这不只是FORTRAN相关行为（虽然只有ifort编译code给出上述错误）

 ＃包括LT＆;＆unistd.h中GT;
＃包括LT＆; SYS / types.h中＆GT;
＃包括LT＆; SYS / stat.h＆GT;
＃包括LT＆;＆fcntl.h GT;诠释主（）{
  INT N，F;
  F =打开（foo.txt的，O_WRONLY | O_CREAT | O_APPEND，0666）;
  写（F，A \\ N，3）;
  N = lseek的（F，0，SEEK_CUR）;
  ftruncate（F，N）;
  写（F，B \\ N，3）;
}
 

但奇怪的是，当我运行code上面我的笔记本电脑确实在意料之中的事情，给我运行程序AB 所以逻辑上似乎 lseek的使我使用群集上的问题，但我知道一点吧（我不知道如何解决它），所以我问在这里。

  

为什么会出现这种情况，最重要的是：？有没有办法解决这个问题。

这是很烦人的，因为为了避免它，我需要做一个临时文件和管道从运行到这个输出，然后condencate与临时文件的原始文件之后，以获得所需的输出为 foo.txt的。

系统规格在那里我有这个问题：

  ifort 14.0.2 20140120
GCC 4.4.7 20120313
Linux集群2.6.32-431.29.2.el6.x86_64＃1 SMP太阳7月27日15时55分46秒美国东部时间2014年的x86_64 x86_64的x86_64的GNU / Linux的
 

和它的工作。

  GCC 4.8.5
达尔文笔记本13.4.0 Darwin内核版本13.4.0：太阳8月17日19点50分11秒PDT 2014年;根：XNU-2422.115.4〜1 / RELEASE_X86_64 x86_64的
 

解决方案

我的Fortran语言很生疏，但我认为，在默认情况下Fortran语言将每个写作为一个单独的记录，这是导致新线（和寻求 - 这似乎是一个越野车对我来说，太）。

如果你想用晚餐preSS这种行为，这可能是一个有点接近你要找的是什么：

 程序的主
写（*，100，事先='无'）'A'
写（*，100，事先='无'）'B'
100格式（A）
程序结束
 

I have a strange bug when appending to file in bash where text gets deleted even though I’m telling it to append.

I have a bash-script where I do the following

echo "Run program" > foo.txt
./fortran_program >> foo.txt

The program fortran_program (a stripped down version of it that gives the same result) is:

program main 
write(*,*) 'A'
write(*,*) 'B'
end program

This should give me Run program AB in foo.txt after the execution. However I find Run B instead so the original text has been overwritten and A is not included.

I did strace -f ./script and found the following output that looks relevant:

...
[pid 36681] open("foo.txt", O_WRONLY|O_CREAT|O_APPEND, 0666) = 3
[pid 36681] dup2(3, 1)                  = 1
[pid 36681] close(3)                    = 0
...
[pid 34260] write(1, " A\n", 3)         = 3
[pid 34260] lseek(1, 0, SEEK_CUR)       = 3
[pid 34260] ftruncate(1, 3)             = 0
[pid 34260] write(1, " B\n", 3)         = 3
...

From my very limited understanding of this it seems that A is written to the file, then lseek is called which gives that the file is only 3 characters long (same length as the first string I try to write) and then ftruncate truncates it at 3 characters (i.e. Run) and then writes B to the file.

The following program in c (echo "Run program" > foo.txt; ./c_program) reproduces the behaviour so it seems it is not just fortran related (though only ifort compiled code gives the error above)

#include<unistd.h>
#include<sys/types.h>
#include<sys/stat.h>
#include<fcntl.h>

int main(){
  int n, f;
  f = open("foo.txt", O_WRONLY | O_CREAT | O_APPEND, 0666);
  write(f," A\n",3);
  n = lseek(f,0,SEEK_CUR);
  ftruncate(f,n);
  write(f," B\n",3);
}

The strange thing is that when I run the code above on my laptop it does the expected thing and gives me Run Program AB so it seems logical that lseek causes the problem on the cluster I use, but I know to little of it (and I have no idea how to fix it) so I'm asking it here.

Why does this happen and most importantly: is there a way to fix this?

This is quite annoying since to avoid it I need to make a temp file and pipe the output from the run into this and then condencate the original file with the temp-file afterwards to get the desired output to foo.txt.

System specifications where I have the problem:

ifort 14.0.2 20140120
gcc 4.4.7 20120313
Linux cluster 2.6.32-431.29.2.el6.x86_64 #1 SMP Sun Jul 27 15:55:46 EDT 2014 x86_64 x86_64 x86_64 GNU/Linux

and where it's working

gcc 4.8.5
Darwin laptop 13.4.0 Darwin Kernel Version 13.4.0: Sun Aug 17 19:50:11 PDT 2014; root:xnu-2422.115.4~1/RELEASE_X86_64 x86_64

解决方案

My Fortran is very rusty, but I think that by default Fortran treats each write as a separate record, which is resulting in the new lines (and the seeks? - which seems a buggy to me, too).

If you want to suppress that behavior, this might be a little closer to what you're looking for:

program main 
write(*,100, advance='no') 'A'
write(*,100, advance='no') 'B'
100 format (A)
end program

这篇关于追加输出文件的时候奇怪的错误的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！