Retry read() when getting EAGAIN

When download is requested from new client, client DB including several tables is created.
This often takes long time on RPI4 and makes IPC error(EAGAIN) in the end.

Change-Id: Ib28dcc22e929d11fe61e5f1690f7da64ee5465b8

diff --git a/packaging/download-provider.spec b/packaging/download-provider.spec
index 7494d34..cda6917 100755 (executable)
--- a/packaging/download-provider.spec
+++ b/packaging/download-provider.spec
@@ -1,6 +1,6 @@
 Name:       download-provider
 Summary:    Download the contents in background
-Version:    2.3.1
+Version:    2.3.2
 Release:    0
 Group:      Development/Libraries
 License:    Apache-2.0

diff --git a/provider/download-provider-ipc.c b/provider/download-provider-ipc.c
index dfe587e..d1e8eea 100644 (file)
--- a/provider/download-provider-ipc.c
+++ b/provider/download-provider-ipc.c
@@ -101,7 +101,7 @@ dp_ipc_fmt *dp_ipc_get_fmt(int sock)
        memset(ipc_info, 0x00, sizeof(dp_ipc_fmt));
 
        int errorcode = DP_ERROR_NONE;
-       int tryagain = 3;
+       int tryagain = 5;
 
        do {
                ssize_t recv_size = read(sock, ipc_info, sizeof(dp_ipc_fmt));
@@ -111,7 +111,8 @@ dp_ipc_fmt *dp_ipc_get_fmt(int sock)
                } else {
                        errorcode = DP_ERROR_NONE;
                }
-       } while (sock >= 0 && (errorcode == DP_ERROR_IO_EINTR) && (--tryagain > 0));
+       } while (sock >= 0 && (errorcode == DP_ERROR_IO_EAGAIN ||
+               errorcode == DP_ERROR_IO_EINTR) && (--tryagain > 0));
 
        if (errorcode != DP_ERROR_NONE) {
                TRACE_ERROR("Failed to read sock[%d]", sock);