Backport 37ed86f (ruby#10248)

nurse · web-flow · commit 577f9c7a8334 · 2024-03-14T07:53:14.000Z
merge revision(s) 37ed86f: [Backport #--ticket=20161] Fix memory leak in regexp grapheme clusters [Bug #20161] The cc->mbuf gets overwritten, so we need to free it to not leak memory. For example: str = "hello world".encode(Encoding::UTF_32LE) 10.times do 1_000.times do str.grapheme_clusters end puts `ps -o rss= -p #{$$}` end Before: 15536 15760 15920 16144 16304 16480 16640 16784 17008 17280 After: 15584 15584 15760 15824 15888 15888 15888 15888 16048 16112 --- regparse.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/regparse.c b/regparse.c
@@ -6105,7 +6105,8 @@ node_extended_grapheme_cluster(Node** np, ScanEnv* env)
           R_ERR(add_code_range(&(cc->mbuf), env, 0x000A, 0x000A)); /* CR */
           R_ERR(add_code_range(&(cc->mbuf), env, 0x000D, 0x000D)); /* LF */
           R_ERR(not_code_range_buf(env->enc, cc->mbuf, &inverted_buf, env));
-          cc->mbuf = inverted_buf; /* TODO: check what to do with buffer before inversion */
+          bbuf_free(cc->mbuf);
+          cc->mbuf = inverted_buf;
 
           env->warnings_flag &= dup_not_warned; /* TODO: fix false warning */
         }
diff --git a/version.h b/version.h
@@ -11,7 +11,7 @@
 # define RUBY_VERSION_MINOR RUBY_API_VERSION_MINOR
 #define RUBY_VERSION_TEENY 0
 #define RUBY_RELEASE_DATE RUBY_RELEASE_YEAR_STR"-"RUBY_RELEASE_MONTH_STR"-"RUBY_RELEASE_DAY_STR
-#define RUBY_PATCHLEVEL 12
+#define RUBY_PATCHLEVEL 13
 
 #include "ruby/version.h"
 #include "ruby/internal/abi.h"

Original file line number	Diff line number	Diff line change
`@@ -6105,7 +6105,8 @@ node_extended_grapheme_cluster(Node** np, ScanEnv* env)`
`6105`	`6105`	`R_ERR(add_code_range(&(cc->mbuf), env, 0x000A, 0x000A)); /* CR */`
`6106`	`6106`	`R_ERR(add_code_range(&(cc->mbuf), env, 0x000D, 0x000D)); /* LF */`
`6107`	`6107`	`R_ERR(not_code_range_buf(env->enc, cc->mbuf, &inverted_buf, env));`
`6108`		`- cc->mbuf = inverted_buf; /* TODO: check what to do with buffer before inversion */`
	`6108`	`+ bbuf_free(cc->mbuf);`
	`6109`	`+ cc->mbuf = inverted_buf;`
`6109`	`6110`
`6110`	`6111`	`env->warnings_flag &= dup_not_warned; /* TODO: fix false warning */`
`6111`	`6112`	`}`