FIX: improved handling of virtual files on POSIX systems

resolves: Oldes/Rebol-issues#2303

Author: Oldes

src/core/p-file.c

@@ -3,7 +3,7 @@
 **  REBOL [R3] Language Interpreter and Run-time Environment
 **
 **  Copyright 2012 REBOL Technologies
-**  Copyright 2012-2022 Rebol Open Source Contributors
+**  Copyright 2012-2024 Rebol Open Source Contributors
 **  REBOL is a trademark of REBOL Technologies
 **
 **  Licensed under the Apache License, Version 2.0 (the "License");
@@ -275,15 +275,24 @@ REBINT Mode_Syms[] = {
 {
 	REBSER *ser;
 	REBVAL *ds = DS_RETURN;
+	REBINT res;
 
+resize:
 	// Allocate read result buffer:
 	ser = Make_Binary(len);
 	Set_Series(REB_BINARY, ds, ser); //??? what if already set?
 
 	// Do the read, check for errors:
 	file->data = BIN_HEAD(ser);
 	file->length = len;
-	if (OS_DO_DEVICE(file, RDC_READ) < 0) return; // Trap_Port(RE_READ_ERROR, port, file->error);
+	res = OS_DO_DEVICE(file, RDC_READ);
+	if (res == -RFE_RESIZE_SERIES) {
+		// We are reading virtual file where the size was initialy reported as 0,
+		// but now we have the real size calculated, so allocate the buffer again.
+		len = file->file.size;
+		goto resize;
+	}
+	if (res < 0) return; // Trap_Port(RE_READ_ERROR, port, file->error);
 	// Not throwing the error from here!
 	// We may want to close the port before error reporting.
 	// It is passed above in the file->error

src/include/reb-file.h

@@ -3,6 +3,7 @@
 **  REBOL [R3] Language Interpreter and Run-time Environment
 **
 **  Copyright 2012 REBOL Technologies
+**  Copyright 2012-2024 Rebol Open Source Developers
 **  REBOL is a trademark of REBOL Technologies
 **
 **  Licensed under the Apache License, Version 2.0 (the "License");
@@ -53,6 +54,7 @@ enum {
 	RFE_BAD_READ,		// Read failed (general)
 	RFE_BAD_WRITE,		// Write failed (general)
 	RFE_DISK_FULL,		// No space on target volume
+	RFE_RESIZE_SERIES,  // Used on Posix to report, that the target series must be resized
 };
 
 #define MAX_FILE_NAME 1022

src/os/posix/dev-file.c

@@ -3,6 +3,7 @@
 **  REBOL [R3] Language Interpreter and Run-time Environment
 **
 **  Copyright 2012 REBOL Technologies
+**  Copyright 2012-2024 Rebol Open Source Developers
 **  REBOL is a trademark of REBOL Technologies
 **
 **  Licensed under the Apache License, Version 2.0 (the "License");
@@ -432,7 +433,22 @@ static int Get_File_Info(REBREQ *file)
 	return DR_DONE;
 }
 
-
+// Resolves real size of virtual files (like /proc/cpuinfo)
+static size_t get_virtual_file_size(const char *filepath) {
+	#define BUFFER_SIZE 4096
+	char buffer[BUFFER_SIZE];
+	size_t size = 0;
+	int file = open(filepath, O_RDONLY, S_IREAD);
+	if (file) {
+		while (1) {
+			size_t bytesRead = read(file, buffer, BUFFER_SIZE);
+			if (bytesRead == 0) break;
+			size += bytesRead;
+		}
+		close(file);
+	}
+	return size;
+}
 /***********************************************************************
 **
 */	DEVICE_CMD Read_File(REBREQ *file)
@@ -463,16 +479,31 @@ static int Get_File_Info(REBREQ *file)
 		if (!Seek_File_64(file)) return DR_ERROR;
 	}
 
-	// printf("read %d len %d\n", file->id, file->length);
-	num_bytes = read(file->id, file->data, file->length);
-	if (num_bytes < 0) {
-		file->error = -RFE_BAD_READ;
-		return DR_ERROR;
-	} else {
-		file->actual = num_bytes;
-		file->file.index += file->actual;
+	// virtual files on Posix report its size as 0, so try to resolve the real one
+	// but only in case, when user did not set /part
+	if (file->file.size == 0 && file->length == 0) {
+		file->file.size = get_virtual_file_size(file->file.path);
+		if (file->file.size > 0 && file->length < file->file.size) {
+			file->error = -RFE_RESIZE_SERIES;
+			return DR_ERROR;
+		}
 	}
 
+	// printf("read %d len %d\n", file->id, file->length);
+	file->actual = 0;
+	// Using the loop, because the reading may be done in chunks!
+	while (1) {
+		num_bytes = read(file->id, file->data + file->actual, file->length - file->actual);
+		if (num_bytes == 0) break;
+		if (num_bytes < 0) {
+			file->error = -RFE_BAD_READ;
+			return DR_ERROR;
+		}
+		file->actual += num_bytes;
+		// stop in case that we have enough data (requested just part of it)
+		if (file->actual >= file->length) break;
+	}
+	file->file.index += file->actual;
 	return DR_DONE;
 }
 

src/tests/units/port-test.r3

@@ -582,14 +582,21 @@ if system/platform = 'Windows [
 if exists? %/proc/cpuinfo [
 	--test-- "Reading from /proc files on Linux"
 	;@@ https://github.com/Oldes/Rebol-issues/issues/2303
+		;; reading complete file
 		--assert all [
-			not error? info: try [read/string %/proc/cpuinfo]
-			empty? info ;; empty, because to read this type of file, the size must be specified!
+			not error? info: try [read %/proc/cpuinfo]
+			0 < len: length? info
+			print to string! info
 		]
+		;; test when requested longer part of the virtual file
 		--assert all [
-			not error? info: try [read/string/part %/proc/cpuinfo 10000]
-			print info
-			0 < length? info
+			not error? info: try [read/part %/proc/cpuinfo len + 1000]
+			len == length? info
+		]
+		;; test when requested just a short part of the virtual file
+		--assert all [
+			not error? info: try [read/part %/proc/cpuinfo 10]
+			10 == length? info
 		]
 ]
 	--test-- "Reading an empty file"