Commitc981879

committed

Don't bother to set sockaddr_un.sun_len.

It's not necessary to fill in sun_len when calling bind() or connect(),on all known systems that have it.Discussion:https://postgr.es/m/2781112.1644819528%40sss.pgh.pa.us

1 parentf972ec5 commitc981879Copy full SHA for c981879

File tree

-14

lines changed

-14

lines changed

Lines changed: 0 additions & 14 deletions

Original file line number	Diff line number	Diff line change
`@@ -218,20 +218,6 @@ getaddrinfo_unix(const char path, const struct addrinfo hintsp,`
`218`	`218`	`aip->ai_addrlen= offsetof(structsockaddr_un,sun_path)+strlen(path);`
`219`	`219`	`}`
`220`	`220`
`221`		`-/*`
`222`		`- * The standard recommendation for filling sun_len is to set it to the`
`223`		`- * struct size (independently of the actual path length). However, that`
`224`		`- * draws an integer-overflow warning on AIX 7.1, where sun_len is just`
`225`		`- * uint8 yet the struct size exceeds 255 bytes. It's likely that nothing`
`226`		`- * is paying attention to sun_len on that platform, but we have to do`
`227`		`- * something with it. To suppress the warning, clamp the struct size to`
`228`		`- * what will fit in sun_len.`
`229`		`- */`
`230`		`-#ifdefHAVE_STRUCT_SOCKADDR_SA_LEN`
`231`		`-unp->sun_len=Min(sizeof(structsockaddr_un),`
`232`		`- ((size_t)1 << (sizeof(unp->sun_len)*BITS_PER_BYTE))-1);`
`233`		`-#endif`
`234`		`-`
`235`	`221`	`return0;`
`236`	`222`	`}`
`237`	`223`

Comments

(0)