If multiple tcl scripts are running in the same directory, they can crash if one tries to auto_mkindex at the same exact time as another.
How can I prevent this properly? I do not want to just place catch around auto_mkindex, nor do I want to implement a semaphore system for this simple problem.
Why would you be building the tclIndex files at the same time in the first place? That's a step that I would expect as part of installation (i.e., something done once as a special action) and not as part of operation (i.e., many times, in parallel potentially). If it's part of installation, it's entirely your own problem if you try to run the code while you're installing it.
I also wouldn't tend to use tclIndex for anything shared between applications, as that's optimized for simple scripts. Shared components are better off made into packages, especially as they're versioned entities, and they have their own indexing mechanism (the pkgIndex.tcl). (Having the same version of the same package installed twice in such a way that things interfere… well, that wouldn't be sensible, would it?)
I am working on a configuration program - it isn't autoconf, but I'm trying (as much as possible) to get it so that ./configure files that use it can be interfaced in a similar manner to those that are made with autoconf -- and that means (as much as possible) supporting the same variable options.
Only there's one option that makes no sense to me. I mean, yes, I am fully clear on what the option means, I just can't conceive of a single scenario in which someone would be well-advised to use that option - except for one scenario in which I'm equally curious why ./configure scripts can't auto-detect the information it would provide.
The option I am referring to is the "--srcdir" option. The reason it so befuddles me is that the only scenario I can imagine in which the source-code files won't be in your present-working-directory (or relative to your present-working-directory as the configure script is programmed to expect) is if the "configure" script itself isn't in your present-working-directory ---- and in that one scenario, I really am unable to imagine why the "configure" script can't extrapolate the source-directory from the name it is invoked by - and instead has to have that --srcdir option to give it that information.
For example, let's say your program's source-code is located in the "awesome/software" directory. That means that, from where you are, the "configure" script would be "awesome/software/configure". Why can't the "configure" script deduce that the source-directory is "awesome/software" just from the fact that it is invoked by the name "awesome/software/configure", and instead require me to add a separate command-line option of: --srcdir=awesome/software
And if this is not the kind of scenario where one would need to specify the --srcdir option (or if it is not the only kind of such scenario) can someone describe to me any other kind of scenario where the person installing a program would be well-advised to alter the "srcdir" variable from it's default?
The option I am referring to is the "--srcdir" option. ...
the only scenario I can imagine in which the source-code files won't be in your present-working-directory (or relative to your present-working-directory as the configure script is programmed to expect) is if the "configure" script itself isn't in your present-working-directory
Right.
and in that one scenario, I really am unable to imagine why the "configure" script can't extrapolate the source-directory from the name it is invoked by - and instead has to have that --srcdir option to give it that information.
I'm not sure it's required. The configure script will attempt to guess the location of srcdir:
# Find the source files, if location was not specified.
if test -z "$srcdir"; then
ac_srcdir_defaulted=yes
# Try the directory containing this script, then the parent directory.
...
So if it's in neither of those places, this will fail, hence the need for --srcdir. Maybe this is (was?) needed where there's some kind of performance differential where the sources are stored on a "slow" drive and the build happens on a "fast" drive, and configure seems to run faster on the "fast" drive so it needs to be there as well...
At any rate, --srcdir is just a variable assignment, so it's not hard to do.
Why can't the "configure" script deduce that the source-directory is "awesome/software" just from the fact that it is invoked by the name "awesome/software/configure"
The configure source seems to do that without specifying --srcdir, but I have not tried it.
This may be a stupid question, as most of my programming consists of one-man scientific computing research prototypes and developing relatively low-level libraries. I've never programmed in the large in an enterprise environment before. I've always wondered, what are the main things that logging libraries make substantially easier than just using good old fashioned print statements or file output, simple programming logic and a few global variables to determine how verbosely things get logged? How do you know when a few print statements or some basic file output ain't gonna cut it and you need a real logging library?
Logging helps debug problems especially when you move to production and problems occur on people's machines you can't control. Best laid plans never survive contact with the enemy, and logging helps you track how that battle went when faced with real world data.
Off the shel logging libraries are easy to plug in and play in less than 5 minutes.
Log libraries allow for various levels of logging per statement (FATAL, ERROR, WARN, INFO, DEBUG, etc).
And you can turn up or down logging to get more of less information at runtime.
Highly threaded systems help sort out what thread was doing what. Log libraries can log information about threads, timestamps, that ordinary print statements can't.
Most allow you to turn on only portions of the logging to get more detail. So one system can log debug information, and another can log only fatal errors.
Logging libraries allow you to configure logging through an external file so it's easy to turn on or off in production without having to recompile, deploy, etc.
3rd party libraries usually log so you can control them just like the other portions of your system.
Most libraries allow you to log portions or all of your statements to one or many files based on criteria. So you can log to both the console AND a log file.
Log libraries allow you to rotate logs so it will keep several log files based on many different criteria. Say after the log gets 20MB rotate to another file, and keep 10 log files around so that log data is always 100MB.
Some log statements can be compiled in or out (language dependent).
Log libraries can be extended to add new features.
You'll want to start using a logging libraries when you start wanting some of these features. If you find yourself changing your program to get some of these features you might want to look into a good log library. They are easy to learn, setup, and use and ubiquitous.
There are used in environments where the requirements for logging may change, but the cost of changing or deploying a new executable are high. (Even when you have the source code, adding a one line logging change to a program can be infeasible because of internal bureaucracy.)
The logging libraries provide a framework that the program will use to emit a wide variety of messages. These can be described by source (e.g. the logger object it is first sent to, often corresponding to the class the event has occurred in), severity, etc.
During runtime the actual delivery of the messaages is controlled using an "easily" edited config file. For normal situations most messages may be obscured altogether. But if the situation changes, it is a simpler fix to enable more messages, without needing to deploy a new program.
The above describes the ideal logging framework as I understand the intention; in practice I have used them in Java and Python and in neither case have I found them worth the added complexity. :-(
They're for logging things.
Or more seriously, for saving you having to write it yourself, giving you flexible options on where logs are store (database, event log, text file, CSV, sent to a remote web service, delivered by pixies on a velvet cushion) and on what is logged at runtime, rather than having to redefine a global variable and then recompile.
If you're only writing for yourself then it's unlikely you need one, and it may introduce an external dependency you don't want, but once your libraries start to be used by others then having a logging framework in place may well help your users, and you, track down problems.
I know that a logging library is useful when I have more than one subsystem with "verbose logging," but where I only want to see that verbose data from one of them.
Certainly this can be achieved by having a global log level per subsystem, but for me it's easier to use a "system" of some sort for that.
I generally have a 2D logging environment too; "Info/Warning/Error" (etc) on one axis and "AI/UI/Simulation/Networking" (etc) on the other. With this I can specify the logging level that I care about seeing for each subsystem easily. It's not actually that complicated once it's in place, indeed it's a lot cleaner than having if my_logging_level == DEBUG then print("An error occurred"); Plus, the logging system can stuff file/line info into the messages, and then getting totally fancy you can redirect them to multiple targets pretty easily (file, TTY, debugger, network socket...).
File systems are volatile. This means that you can't trust the result of one operation to still be valid for the next one, even if it's the next line of code. You can't just say if (some file exists and I have permissions for it) open the file, and you can't say if (some file does not exist) create the file. There is always the possibility that the result of your if condition will change in between the two parts of your code. The operations are distinct: not atomic.
To make matters worse, the nature of the problem means that if you're tempted to make this check, odds are you're already worried or aware that something you don't control is likely to happen to the file. The nature of development environments make this event less likely to happen during your testing and very difficult to reproduce. So not only do you have a bug, but the bug won't show up while testing.
Therefore under normal circumstances the best course of action is to not even try to check if a file or directory exists. Instead, put your development time into handling exceptions from the file system. You have to handle these exceptions anyway, so this is a much better use of your resources. Even though exceptions are slow, checking the existence of a file requires an extra trip to disk, and disk access is much slower. I even have a well-voted answer to this effect in another question.
But I'm having some doubts. In .Net, for example, if that's really always true, the .Exists() methods wouldn't be in the API in the first place. Also consider scenarios where you expect your program to need to the create file. The first example that comes to mind is for a desktop application. This application installs a default user-config file to it's home directory, and the first time each user starts the application it copies this file to that user's application data folder. It expects the file not to exist on that first startup.
So when is it acceptable to check in advance for the existence (or other attributes, like size and permissions) of a file? Is expecting failure rather than success on the first attempt a good enough rule of thumb?
The File.Exists method exists primarily for testing for the existence of a file when you do not intend to open the file. For example testing for the existence of a locking file whose very existence tells you something but whose contents are immaterial.
If you are going to open the file then you will need to handle any exception regardless of the results of any prior calls to File.Exists. So, in general, there is no real value in calling it in these circumstances. Just use the appropriate FileMode enumeration value in your open method and handle any exceptions, as simple as that.
EDIT: Even though this is couched in terms of the .Net API, it is based on the underlying system API. Both Windows and Unix have system calls (i.e. CreateFile) that use the equivalent of the FileMode enumeration. In fact in .Net (or Mono) the FileMode value is just passed through to the underlying system call.
As a general policy, methods like File.Exists, or properties like WeakReference.Alive or SomeConcurrentQueue.Count are not useful as a means of ensuring that a "good" state exists, but can be useful as a means of determining that a "bad" state exists without doing any unnecessary (and possibly counterproductive) work. Such situations may arise in many scenarios involving locks (and files, since they often include locks). Because all routines that need to lock on a set of resources should, whenever practical, always acquire locks on those resources in a consistent order, it may be necessary to acquire a lock on one resource which is expected to exist before acquiring a resource which may or may not exist. In such a scenario, while it's impossible to avoid the possibility that one might lock the first resource, fail to acquire the second, and then release the first lock without having done any useful work with it, checking for the existence of the second resource before acquiring the lock on the first would minimize unnecessary and useless effort.
It depends on your requirements, but one way is to try to obtain an exclusive open file handle, with some sort of retry mechanism. Once you have that handle, it's going to be hard (or impossible) for another process to delete (or move) that file.
I've used code in .NET similiar to the following to obtain an exclusive file handle, where I expect some other process to be possibly writing the file:
FileInfo fi = new FileInfo(fullFilePath);
int attempts = maxAttempts;
do
{
try
{
// Asking to open for reading with exclusive access...
fs = fi.Open(FileMode.Open, FileAccess.Read, FileShare.None);
}
// Ignore any errors...
catch {}
if (fs != null)
{
break;
}
else
{
Thread.Sleep(100);
}
}
while (--attempts > 0);
One example: You may be able to check for existence of files which you are unable to open (due to, for example, permissions).
Another, possibly better example: You want to check for the existence of a Unix device file. But definitely do not open it; opening it has side effects (e.g., open/close /dev/st0 will rewind the tape)
In *nix environment a well established method for checking if another copy of the program is already running is to create a lock file. So the check for file existence is used to verify this.
I'd only check it if I expect it to be missing (e.g. the application settings) and only if I have to read the file.
If I have to write to the file, it's either a logfile (so I can just append to it or create a new one) or I replace the contents of it, so I might as well recreate it anyway.
If I expect that the file exists, it would be right that an Exception is thrown. Exception handling should then inform the user or perform recovery. My opinion is that this results in cleaner code.
File protection (i.e. not overwriting (possibly important) files) is different, in that case I'd always check whether a file exists, if the framework doesn't do that for me (think SaveFileDialog)
I think the check makes sense when you want to be sure the file was there in the first place. As you said settings files...if there is a file I will try and merge the existing settings instead of blowing them away.
Other cases would be when a user tells me to do something with a file. Yes I know the openFileDialog will check if a file exists (But this is optional). I vaguely remeber back in VB6 this was not the case, so verifying the file existed that they just told me to use was common.
I'd rather not program by exception.
Edit
I didn't miss the point. You might try and access the file, an exception is thrown and then when you go to create the file, the file was already placed there. Which now causes your exception handling code to go on the fritz. So I guess we could then have an exception handler in our exception handler to catch that the file changed yet again...
I'd rather try and prevent exceptions, not use them to control logic.
Edit
Additionally another time to check for attributes such as size is when your waiting for a file operation to finish, yes you never know for sure but with a good algorithim and depending on the system writting the file you might be able to handle a good deal of cases (Had a system running for five years which watched for small files coming over ftp, and it uses a the same api as the file system watcher, and then starts polling waiting for the file to stop changing, before raising an event that the file is ready to be consumed).
This may be too simplistic, but I would think the primary reason for checking for the existence of a file (hence the existence of .Exists()) would be to prevent unintended overwrites of existing files, not to avoid exceptions caused by attempting to access non-existent nor non-accessible files.
EDIT 2
This was, in fact, too simplistic and I recommend you see Stephen Martin's response.
If you're that concerned about somebody else removing the file, perhaps you should implement some sort of locking system. For instance, I used to work on the code for C-News, a Usenet news server. Since a lot of the things it did could happen asynchronously, it would "lock" a file or a directory by making a temp file, and then hard linking it to a file named "LOCK". If the link failed, it would mean that some other version of the program was writing to that directory, otherwise it was yours and you could do what you like.
The nifty thing about this is that most of the program was written in shell and awk, and this was a very portable locking mechanism. Also, the lock file would contain the PID of the owner, so you could look at the existing lock file to see if the owner was still running.
We have a diagnostic tool that has to gather a set of files, installer log included. Depending on different conditions the installer log can be in one of two folders. Even worse, there can be different versions of the log in both of these folders. How does the tool find the right one?
It's quite simple if you check for existence. If only one is present, grab that file. If two exist, find which has the latest modification time and grab that file. That's just normal way of doing things.
While this is a language-agnostic post, it seems you are talking about .NET. Most systems (.NET and others) have more detailed APIs in order to figure out if the file exists when opening the file.
What you should do is make a call to access the file, as it will typically indicate through some sort of error that the file doesn't exist (if it truly doesn't). In .NET, you would have to go through the P/Invoke layer and use the CreateFile API function. If that function returns an error of ERROR_FILE_NOT_FOUND, then you know that the file does not exist. If it returns successfully, then you have a handle that you can use.
The point here is that it is a somewhat atomic operation, which ultimately is what you are looking for.
Then, with the handle, you can pass it to a FileStream constructor and perform your work on the file.
There are a numbers of possible applications you may well be writing that a simple File.Exists is more than adequate for the job. If it's a config file that only your application will use then you do not need to go so overkill in your exception handling.
Whilst the "flaws" you have pointed out in using this method are all valid, it doesn't mean they are not acceptable flaws for some situations.
A variety of apps include built-in web servers. It's common for them to generate self-signed SSL certificates the first time they start up. A straightforward way to implement this would be to check whether the cert exists on startup, and create it if not.
In theory, it could exist for the check, and not exist later. In that case, we'd get an error when we try to listen, but that can be handled quite easily and is not a big deal.
It's also possible that it doesn't exist for the check, and exists later. In that case, it either gets overwritten with a new cert, or writing the new cert fails, depending on your policy. The first is a little annoying, in terms of the cert change causing some alarm, but also not really critical, especially if you do a bit of logging to indicate what is going on.
And, in practice, both cases are extraordinarily unlikely to ever come up.
Like you pointed out its always important what the program should do if the file is missing. In all my applications the user can always delete the config file and the application will create a new one with default values. No Problem. I also ship my applications without config files.
But users tend to delete files and even files they should not delete like serial keys and template files. I always check for these files because without them the application is unable to run at all. I can not create a new serial key from default.
Whats should happen when the file is missing? You can do a file find or exception handler but the real question is : What will happen when the file is missing? Or how important is the file for the application. I check all the time before I try to access any support files for the app. Additional I do error handling if the file is corrupt and can not be loaded.
I think anytime that you know that the file may or may not exist and you want to perform some alternate action based on the existence of the file, you should do the check because in this case it's not an exceptional condition for the file to not exist. This won't absolve you from having to handle exceptions -- from someone else either removing or creating the file between the check and your open -- but it makes the intent of the program clear and doesn't rely on exception handling to perform flow-control logic.
EDIT: An example might be log rotation on start up.
try
{
if (File.Exists("app.log"))
{
RotateLogs();
}
log = File.Open("app.log", FileMode.CreateNew );
}
catch (IOException)
{
...another writer, perhaps?
}
catch (UnauthorizedAccessException)
{
...maybe I should have used runas?
}
To answer my own question (in part), I want to expand on the example I used: a default config file.
Rather than check if it exists at app startup and try to copy the file if the check fails, the thing to do is always try to copy the file. You just do it in such a way that the copy will fail if the file exists rather than replace an existing file. This way all you need to do is catch and ignore any exception thrown if the copy fails because of an existing file.
Your problem could easily be solved with basic computer science... read up on Semaphores.
(I did not mean to sound like a jerk, I was just pointing you to a simple answer for a common problem).
I think the reason for "Exists" is to determine when files are missing without the need for creating all the OS housekeeping data required to access the file or having exceptions being thrown. So it's a file handling optimisation more than anything else.
For a single file, the saving the "Exists" gives is generally insignificant. If you were checking if a file exists many, many times (for example, searching for #include files) then the saving could be significant.
In .Net, the specification for File.Exists doesn't list any exceptions that the method might throw, unlike for example File.Open which lists nine exceptions, so there's certainly less checking going on in the former.
Even if "Exists" returns true, you still need to handle exceptions when opening the file, as the .Net reference suggests.