JPH03246733A - Distributed program debugging system - Google Patents

Abstract

PURPOSE:To permit designation for conditional equations relating to plural processes to a user and to stop processing or output the value of an arithmetic equation at a time when the minimum number of times of an operation to satisfy the conditional equation is performed by enabling the local control part of each process to communicate the information of a process state with each other when the operation is reproduced. CONSTITUTION:A function to communicate the information relating to its own process 2 with the local control part of another process by the local control part 5 in the reproduction of the operation is provided. When the conditional equations relating to the plural processes are supplied and it is required to operate another process to satisfy the conditional equation of a certain process, the driving requirement of the process is sent to the local control part 11 which controls the operation of the process. The operation of each process 2 can be controlled by a sent driving requirement. Thereby, the designation by the conditional equations relating to the plural processes can be permitted, and stoppage and the output of the arithmetic equation can be performed at a time when the minimum number of times of the operation to satisfy the conditional equation are performed.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a debugging method for testing a program in a system having multiple processes on a target computer system to be run.

[Conventional technology]

Conventionally, in a debugger that targets a single distributed program, the following behavior reproduction method has been used. Regarding this, see 1, for example, T-J, LeBlancand J,
M. Mellor-Crummey: “Deb.
uggingParallel Programsw
ith In5tant Replay, “I
E E E Trans, oncow+put,
, vol, c-36.

&4. pp, 471-480 (April 1987
) is discussed in

The configuration of a general debugger is shown in FIG. 1 with the other process control section 9 removed. In FIG.

System 1 consists of a plurality of processes connected through communication paths. A local control unit 5 is provided for controlling the operation of each of these processes 2. Before the global control unit 14 accepts or accepts a user's debugger command, it once executes the program 3 to be debugged. At that time, communication history management department 7
When the own process issues a message reception command,
The name of the process that sent the message is stored in the second internal storage area.
Store in order as shown in the figure.

After accepting the debugger command, re-execute. That is, program 4 is executed again from the beginning.

At the time of this re-execution, the communication history management unit 7 provides a pointer variable. The pointer variable initially points to the first element of the communication history. During this re-execution, if the next command of a certain process is a receive command, the communication history management unit 7 of that process refers to the element of the communication history pointed to by the pointer variable. The communication history management unit 7 returns NO if the next command is a receive command and the message has not arrived via the communication path 3 from the process with the name pointed to by the pointer, otherwise it returns Y.
Output es.

When the operation control unit 6 receives Yes from the communication history management unit 7, it sends Execute to the process and causes it to execute the next command. If NO is received, Execute is not output and the receive command is not executed until it becomes Yes (that is, a message from the process to be received by the next receive command arrives). Furthermore, if a message arrives from a process other than that (that is, a received message for a subsequent receive command arrives earlier), the message is stored in the internal storage area until the corresponding receive command is executed. In this way, as long as the local control unit 5 operates the process so that the communication history is the same, the same operation will be reproduced no matter how many times it is executed.

[Problem to be solved by the invention]

Many of the debuggers that perform reproduction use breakpoint setting commands, which stop when a user-specified conditional expression is met, and trace output commands, which perform user-specified arithmetic operations when a user-specified conditional expression is met. It provides commands to output the value of an expression. In this type of conventional debugging method, the following operations are performed when reproducing the above commands in response to the commands input by the user.

The conditional expression determination unit 8 determines the conditional expression regarding the own process based on the values of the variables of the own process, the value of the program counter, etc., and outputs True if the conditional expression holds, otherwise outputs False. The operation control unit 6 outputs Yes when the communication history management unit 7 outputs Yes, and the conditional expression determination unit 8 outputs False.
When outputting se, Execute is output to execute the next instruction. While the communication history management section 7 is outputting NO, it does not output anything and waits for it to become Yes. Further, when the conditional expression determining unit 8 outputs True, one conditional expression has been established, so re-execution is stopped, and the user is notified through the area control unit 14 that the conditional expression has been established.

In this way, the local control unit 5 controls the reproduction operation only from the conditional expressions related to its own process, and operates without knowing information about the fulfillment status of the conditional expressions of other processes. Therefore, only logical expressions relating to a single process are allowed as conditional expressions for commands, and specification using conditional expressions relating to multiple processes is not possible. For example, the logical formula related to process 1 and process 2 (V = 1) N (V2 = 1) (V2, V2, each is a variable in process 1 and 2) is considered. At this time, in the conventional debugger, V□=1 is given to the conditional expression determination section of process 1 and v2=1 is given to the conditional expression determination section of process 2 as the conditional expression of the breakpoint and executed.

When process 1 and process 2 each execute the program shown in FIG. 3, in the conventional debugger, the conditional expression is satisfied immediately after process 1 executes instruction 13, so the operation control unit stops re-execution.

Therefore, process 2 cannot re-execute beyond instruction 25 even though v2=1 does not hold.

Therefore, it was necessary for the user to once release the conditional expression V = 1 for process 1 to operate the process, and then repeat the process of setting the conditional expression again. Similarly, with respect to the trace output command, the value of an arithmetic expression when a conditional expression regarding a plurality of processes is satisfied cannot be obtained without user intervention.

Also, when the conditional expression V□=1 for only process 1 is given as a breakpoint conditional expression, in the conventional method, process 1 stops immediately after executing instruction 13, but process 2 runs up to instruction 24, In order to satisfy the conditional expression,
It is not possible to stop the system in the state in which each process has performed its minimum operations, that is, in the system state immediately after the error has occurred. Similarly, with respect to the trace output command, it is not possible to obtain the value of the arithmetic expression at the time when the minimum number of operations have been performed to satisfy the conditional expression.

It is an object of the present invention to allow specification of a conditional expression for multiple processes by combining logical expressions for processes by logical product in the conditional expression for a breakpoint setting command, and to specify a logical expression for a process as a conditional expression for a trace output command. It allows the specification of multiple processes using conditional expressions that are combined by logical product or logical sum, and allows the user to stop when the minimum number of operations are performed to satisfy the conditional expression and output the value of the arithmetic expression. The object of the present invention is to provide a distributed program debugging method that allows debugger commands to be written in a natural manner and allows a user to recognize the state of the system immediately after a bug becomes apparent.

[Means to solve the problem]

In order to achieve the above object, the debugging method of the present invention is characterized by providing a function for the local control unit to communicate information regarding its own process to the local control units of other processes when reproducing operations.

[For production]

In this method, when conditional expressions related to multiple processes are given and it is necessary to operate other processes in order to satisfy the conditional expression related to a certain process, the local control unit that controls the operation of the process is Drive requests can be sent. By controlling the operation of each process based on the sent drive request, the user is allowed to specify conditional expressions regarding multiple processes, and the point in time when the minimum amount of operation is performed to satisfy the conditional expression is determined. I can do it.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a configuration diagram illustrating an embodiment of the present invention. 1
Ij! is the distributed program to be debugged. The system is made up of a plurality of processes connected by communication channels in order to operate the system. Reference numeral 2 indicates each of the processes, 3 indicates a communication channel connecting the processes, and 4 indicates a program to be debugged for each process.

5 is a local control unit that controls the operation of the process.

6 is Execute, which instructs the process to perform an operation.
An operation control unit 7 that outputs a signal saves the communication history during operation, and outputs a signal Y when the next command can be executed when re-executing.
A communication history management unit 8 outputs es and controls the communication history to be the same at the time of re-execution, and a communication history management unit 8 determines whether the conditional expression of the user command is satisfied at the time of re-execution, and when it is true, outputs a signal Tru.
e, and outputs a signal False when the condition is not satisfied; 9 is another process control unit that communicates with the local control unit of another process during re-execution; 10 is another process during re-execution 11 is a communication status management unit that records the communication status with the other process; 11 is another process operation control unit that sends a drive request message to the local control unit of the other process when it is necessary to operate the other process; This drive request analysis unit analyzes the received drive request message and outputs a signal Execute Request when the own process needs to perform an operation for another process. Reference numeral 13 denotes a communication path for mutual communication between the local control units. Reference numeral 14 denotes a global control unit for inputting commands from the user and outputting from the debugger.

The following operations are performed for breakpoint setting commands and trace output commands that have conditional expressions related to multiple processes, in which logical expressions for processes are combined by logical AND.

The conditional expression of the command input by the user is S P 1inS
PL, In...n5Plk (SP13 is Process X i
The global control unit 14 divides this logical formula into logical formulas related to a single process, and stores the divided logical formulas SP□ in the storage area of the conditional expression determination unit 8 of the process ij. Store.

At the time of re-execution, the local control unit 5 operates as follows.

(1) The communication state management unit 10 always records the number of messages sent and received by the own process for each partner process.

(2) The communication history management unit 7 controls to reproduce the communication history as described above. It outputs NO only when the next command is received and no message has arrived from the communication path 3; otherwise it outputs Yes.

(3) The conditional expression determination unit 8 returns False when the logical expression regarding the own process is not satisfied, and True otherwise (when the logical expression is satisfied and when there is no logical expression regarding this process). Output.

(4) The operation control unit 6 re-executes the next instruction of its own process when the following conditions are met.

The conditional expression determination unit 8 outputs F also, the drive request analysis unit 12 outputs Execute Request, and the communication history management unit 7 outputs Yes.

(5) When the other process operation control unit 11 needs to operate another process, that is, the conditional expression determination unit 8 outputs False or the drive request analysis unit 12 outputs Ex
ecute and when the communication history management unit 7 outputs No, the following drive request message is sent to the process shown below through the communication path 13.

For the message transmission process (denoted as Q) to be received by the next command, the number of the next message to be received from Q (
A drive request message consisting of 1 added to the number of messages received so far from Q, which is managed by the communication state management section 10, is sent.

(6) When the drive request analysis unit 12 receives a request drive message from the operation control unit of another process (referred to as j), it performs the following operations.

If the number of messages sent to J stored in the communication state management section 10 is smaller than the message number in the drive request message, an Execute Request is output. Otherwise, nothing is output.

(7) The conditional expression is satisfied when the operation control units of all processes stop operating their own processes, and the entire area control unit 14 detects this. If it is a breakpoint setting command, it will notify the user that it has stopped, and if it is a trace output command, it will query each local control unit for the current value of the variable that appears in the arithmetic expression, and then calculate the value of the arithmetic expression from the answer. After outputting, restart each local control unit.

Logical expressions for processes are combined using logical OR.

The trace output command that has conditional expressions related to multiple processes operates as follows.

SP the conditional expression of the trace output command input by the user.
YO, USP Engineering 2U...USP Engineering. The global control unit 14 divides the conditional expression into logical expressions regarding a single process,
The divided logical expressions SP, , are stored in the storage area of the conditional expression determination unit 8 of the process ij. In addition, the variables that appear in the arithmetic expression to be output are divided by process, and the variable name of process l (
X'1. X"2....,X".

) is stored in the storage area of the local control unit of process 1.

The logical expression for each of the divided single processes is the conditional expression S
P□□nSP□2n... can be considered as the case where =1 in USPik. Therefore, the point where the minimum movement that satisfies the conditions can be found using the same method as in the case of the breakpoint setting command described above.

Therefore, re-execution performs the breakpoint setting command procedure described above for the logical formula SP for each single process. In addition to the actual program counter PC representing the location where re-execution is being performed, the operation control unit 6 creates virtual program counters equal to the number of logical expressions related to the divided single process.

FIG. 4 shows the storage status of a set of program counter and variable values. That is, the program counter PC, for each logical formula SP□, attempts to find the point in time when the minimum operation that satisfies SP□ has been performed.

Except for the operation control unit 6, the local control unit operates in response to breakpoint setting commands (1) (2) (3)
The operation control unit 6 performs (4), (5), and (6) independently for m pieces using the logical formula number j as a parameter, as shown in (4') below.
Perform (7').

(4') At the time of re-execution, the operation control unit 6 sets the current variable x'4. x"2....,X". The value is stored together with the value of the program counter PC in the storage area as shown in FIG.

Here, if the condition (4) is satisfied for the logical formula sp, and the operation control unit 6 for the process i operates the process i, the following (A) or Do (B).

(A) If the value of the program counter PC3 for the logical formula SP□1 matches the value of the actual program counter PC (in the case of SF3 in FIG. 4), the process i is actually re-executed.

(B) If the value of the program counter PC for the logical formula SP,j is smaller than the value of the actual program PC (
In the case of SP1 in FIG. 4, since the next instruction has already been re-executed, it is assumed that the value of the program counter PC, has been increased and re-execution has been performed. However, if the command is a transmission command, instead of actually transmitting a message, a virtual message having a logical formula number as a parameter is transmitted.

(7') When any logical formula (sp, J) is established, the operation control unit 6 of each process retrieves the value of the program counter PC3 for that logical formula from the storage area, the variable X'1. ．． The values of X'□, . . . , x゛ゎ are sent to the area control section 14. The global control unit 14 calculates the value of the arithmetic expression from these values and outputs it. Then continue rerunning.

Note that as a method of communication between processes, there is also a method of using shared memory that can be read and written by multiple processes, but for such systems, operations that write to the shared memory are considered as sending, and operations that read from the shared memory are considered as receiving. and the above method can be applied.

〔Effect of the invention〕

As explained above, in the distributed program debugging method of the present invention, the local control units of each process communicate process state information when reproducing operations, allowing the user to specify conditional expressions regarding multiple processes, and It is possible to stop or output the value of an arithmetic expression when the minimum amount of operation that satisfies is completed.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an embodiment of a debugger according to the present invention, FIG. 2 is a diagram showing the storage status of communication history in the communication history management section 7 of FIG. 1, and FIG. 3 is a diagram showing an example of a program. , FIG. 4 is a diagram showing the storage status of a program counter and a set of variables in the operation control unit 6. DESCRIPTION OF SYMBOLS 1... System, 2... Process, 3... Communication path, 4... Program to be debugged, 5... Local control section, 6... Operation control section, 7.
...Communication history management section, 8...Conditional expression judgment section, 9...
・Other process control section. 10...Communication status management unit, 11...Other process operation control unit. 12... Drive requirement analysis section. 13... Communication path between local control units, 14... Area control unit. 10 cess 1 70 cess 2 Change value zoro no ゛'ram power cow 7f) Ren

Claims (1)

[Claims] (1) For distributed programs in a system where multiple processes each have a built-in program and the processes proceed while communicating with each other, the communication history at the time of execution is saved,
In the distributed program debugging method that reproduces the operation based on the history, there is a communication history management means that saves the communication history during operation and controls reproduction according to the communication history, and a conditional expression specified by the user at the time of reproduction. A conditional expression judgment means for determining, a communication status management means for managing the communication status during reproduction, another process operation control means for issuing drive requests to other processes during reproduction, and an analysis of drive requests from other processes during reproduction. A distributed program debugging method comprising a drive request analysis means and an operation control means for controlling process operation.