From 6c951cb023f52226309f1d7cfdb5b86d35fe7fd7 Mon Sep 17 00:00:00 2001
From: Cesar Douady <cesar.douady@doliam.fr>
Date: Fri, 3 May 2024 14:55:55 +0200
Subject: [PATCH] fix rare crash in presence of lost jobs

---
 src/lmakeserver/backend.cc   | 3 +++
 src/lmakeserver/backend.x.hh | 8 ++++++--
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/src/lmakeserver/backend.cc b/src/lmakeserver/backend.cc
index ca73748a..66ec5989 100644
--- a/src/lmakeserver/backend.cc
+++ b/src/lmakeserver/backend.cc
@@ -19,6 +19,7 @@ namespace Backends {
 		auto it   = Backend::_s_start_tab.find(job) ;
 		if (it==Backend::_s_start_tab.end()) return ;                      // job is dead without waiting for reply, curious but possible
 		Backend::StartEntry const& e = it->second ;
+		if (!e                             ) return ;                      // .
 		try {
 			jmrr.seq_id = e.conn.seq_id ;
 			ClientSockFd fd( e.conn.host , e.conn.port , 3/*n_trials*/ ) ;
@@ -166,6 +167,7 @@ namespace Backends {
 
 	void Backend::_s_wakeup_remote( JobIdx job , StartEntry::Conn const& conn , SigDate const& start_date , JobMngtProc proc ) {
 		Trace trace(BeChnl,"_s_wakeup_remote",job,conn,proc) ;
+		SWEAR(conn.seq_id,job,conn) ;
 		try {
 			ClientSockFd fd(conn.host,conn.port) ;
 			OMsgBuf().send( fd , JobMngtRpcReply(proc,conn.seq_id) ) ; // XXX : straighten out Fd : Fd must not detach on mv and Epoll must take an AutoCloseFd as arg to take close resp.
@@ -510,6 +512,7 @@ namespace Backends {
 			for( auto jit = _s_start_tab.begin() ; jit!=_s_start_tab.end() ;) { // /!\ we erase entries while iterating
 				JobIdx      j = jit->first  ;
 				StartEntry& e = jit->second ;
+				if (!e) { jit++ ; continue ; }
 				if (ri) {
 					if ( e.reqs.size()==1 && e.reqs[0]==ri ) goto Kill ;
 					for( auto it = e.reqs.begin() ; it!=e.reqs.end() ; it++ ) { // e.reqs is a non-sorted vector, we must search ri by hand
diff --git a/src/lmakeserver/backend.x.hh b/src/lmakeserver/backend.x.hh
index ff41b5f5..64ff1191 100644
--- a/src/lmakeserver/backend.x.hh
+++ b/src/lmakeserver/backend.x.hh
@@ -54,6 +54,10 @@ namespace Backends {
 			friend ::ostream& operator<<( ::ostream& , StartEntry const& ) ;
 			struct Conn {
 				friend ::ostream& operator<<( ::ostream& , Conn const& ) ;
+				// accesses
+				bool operator+() const { return seq_id  ; }
+				bool operator!() const { return !+*this ; }
+				// data
 				in_addr_t host     = NoSockAddr ;
 				in_port_t port     = 0          ;
 				SeqId     seq_id   = 0          ;
@@ -63,8 +67,8 @@ namespace Backends {
 			StartEntry(       ) = default ;
 			StartEntry(NewType) { open() ; }
 			// accesses
-			bool operator+() const { return conn.seq_id ; }
-			bool operator!() const { return !+*this     ; }
+			bool operator+() const { return +conn   ; }
+			bool operator!() const { return !+*this ; }
 			bool useful   () const ;
 			// services
 			void open() {