Commitde6428a

committed

Avoid another valgrind complaint about write() of uninitalized bytes.

Peter Geoghegan, per buildfarm member skink and Andres FreundDiscussion:http://postgr.es/m/20180221053426.gp72lw67yfpzkw7a@alap3.anarazel.de

1 parent9a5c4f5 commitde6428aCopy full SHA for de6428a

File tree

+12

-0

lines changed

+12

-0

lines changed

Lines changed: 12 additions & 0 deletions

Original file line number	Diff line number	Diff line change
`@@ -739,6 +739,18 @@ LogicalTapeRewindForRead(LogicalTapeSet *lts, int tapenum, size_t buffer_size)`
`739`	`739`	`*/`
`740`	`740`	`if (lt->dirty)`
`741`	`741`	`{`
	`742`	`+/*`
	`743`	`+ * As long as we've filled the buffer at least once, its contents`
	`744`	`+ * are entirely defined from valgrind's point of view, even though`
	`745`	`+ * contents beyond the current end point may be stale. But it's`
	`746`	`+ * possible - at least in the case of a parallel sort - to sort`
	`747`	`+ * such small amount of data that we do not fill the buffer even`
	`748`	`+ * once. Tell valgrind that its contents are defined, so it`
	`749`	`+ * doesn't bleat.`
	`750`	`+ */`
	`751`	`+VALGRIND_MAKE_MEM_DEFINED(lt->buffer+lt->nbytes,`
	`752`	`+lt->buffer_size-lt->nbytes);`
	`753`	`+`
`742`	`754`	`TapeBlockSetNBytes(lt->buffer,lt->nbytes);`
`743`	`755`	`ltsWriteBlock(lts,lt->curBlockNumber, (void*)lt->buffer);`
`744`	`756`	`}`

Comments

(0)