pcre2test: memory reports only compiled memory usage for code/data (#418

)

Since 05aafb2 (Implement pcre2_set_max_pattern_compiled_length() and set
this limit in the fuzzer, 2024-04-24), the memory modifier has reported
the full size of the allocated "code" returned by `pcre2_compile`.

Problem is that the size of the structure used to hold that in memory also
depends on the platform ABI and even alignment by the compiler, and has
been therefore fragile to compare.

Revert to reporting only the additional memory that `pcre2_compile()` will
use for the compiled pattern (including any data tables) and make sure
that the limit provided with `pcre2_set_max_pattern_compiled_length()`
also avoid the internal struct overhead.

Fixes: #415

doc/pcre2_set_max_pattern_compiled_length.3

@@ -1,4 +1,4 @@
-.TH PCRE2_SET_MAX_PATTERN_COMPILED_LENGTH 3 "24 April 2024" "PCRE2 10.44"
+.TH PCRE2_SET_MAX_PATTERN_COMPILED_LENGTH 3 "8 Jun 2024" "PCRE2 10.45"
 .SH NAME
 PCRE2 - Perl-compatible regular expressions (revised API)
 .SH SYNOPSIS
@@ -15,9 +15,9 @@ PCRE2 - Perl-compatible regular expressions (revised API)
 .rs
 .sp
 This function sets, in a compile context, the maximum size (in bytes) for the
-memory needed to hold the compiled version of a pattern that is compiled with
-this context. The result is always zero. If a pattern that is passed to
-\fBpcre2_compile()\fP with this context needs more memory, an error is
+memory needed to hold the compiled version of a pattern that is using this
+context. The result is always zero. If a pattern that is passed to
+\fBpcre2_compile()\fP referencing this context needs more memory, an error is
 generated. The default is the largest number that a PCRE2_SIZE variable can
 hold, which is effectively unlimited.
 .P

src/pcre2_compile.c

@@ -10608,8 +10608,7 @@ block for storing the compiled pattern and names table. Integer overflow should
 no longer be possible because nowadays we limit the maximum value of
 cb.names_found and cb.name_entry_size. */
 
-re_blocksize = sizeof(pcre2_real_code) +
-  CU2BYTES(length +
+re_blocksize = CU2BYTES(length +
   (PCRE2_SIZE)cb.names_found * (PCRE2_SIZE)cb.name_entry_size);
 
 if (re_blocksize > ccontext->max_pattern_compiled_length)
@@ -10618,6 +10617,7 @@ if (re_blocksize > ccontext->max_pattern_compiled_length)
   goto HAD_CB_ERROR;
   }
 
+re_blocksize += sizeof(pcre2_real_code);
 re = (pcre2_real_code *)
   ccontext->memctl.malloc(re_blocksize, ccontext->memctl.memory_data);
 if (re == NULL)

src/pcre2test.c

@@ -4397,7 +4397,7 @@ static void
 show_memory_info(void)
 {
 uint32_t name_count, name_entry_size;
-PCRE2_SIZE size, cblock_size;
+PCRE2_SIZE size, cblock_size, data_size;
 
 /* One of the test_mode values will always be true, but to stop a compiler
 warning we must initialize cblock_size. */
@@ -4417,18 +4417,19 @@ if (test_mode == PCRE32_MODE) cblock_size = sizeof(pcre2_real_code_32);
 (void)pattern_info(PCRE2_INFO_NAMECOUNT, &name_count, FALSE);
 (void)pattern_info(PCRE2_INFO_NAMEENTRYSIZE, &name_entry_size, FALSE);
 
-/* The uint32_t variables are cast before multiplying to stop code analyzers
-grumbling about potential overflow. */
+/* The uint32_t variables are cast before multiplying to avoid potential
+ integer overflow. */
+data_size = (PCRE2_SIZE)name_count * (PCRE2_SIZE)name_entry_size * (PCRE2_SIZE)code_unit_size;
 
-fprintf(outfile, "Memory allocation - compiled block : %" SIZ_FORM "\n", size);
-fprintf(outfile, "Memory allocation - code portion   : %" SIZ_FORM "\n", size -
-  (PCRE2_SIZE)name_count * (PCRE2_SIZE)name_entry_size * (PCRE2_SIZE)code_unit_size -
-  cblock_size);
+fprintf(outfile, "Memory allocation - code size : %" SIZ_FORM "\n", size -
+  cblock_size - data_size);
+if (data_size != 0)
+  fprintf(outfile, "Memory allocation - data size : %" SIZ_FORM "\n", data_size);
 
 if (pat_patctl.jit != 0)
   {
   (void)pattern_info(PCRE2_INFO_JITSIZE, &size, FALSE);
-  fprintf(outfile, "Memory allocation - JIT code       : %" SIZ_FORM "\n", size);
+  fprintf(outfile, "Memory allocation - JIT code  : %" SIZ_FORM "\n", size);
   }
 }