diff options
| -rw-r--r-- | generic/tclIOGT.c | 41 |
1 files changed, 18 insertions, 23 deletions
diff --git a/generic/tclIOGT.c b/generic/tclIOGT.c index 6cc33eb..9c4347d 100644 --- a/generic/tclIOGT.c +++ b/generic/tclIOGT.c @@ -683,39 +683,34 @@ TransformInputProc( read = Tcl_ReadRaw(downChan, buf, toRead); if (read < 0) { - /* - * Report errors to caller. EAGAIN is a special situation. If we - * had some data before we report that instead of the request to - * re-try. - */ - int error = Tcl_GetErrno(); + if (Tcl_InputBlocked(downChan) && (gotBytes > 0)) { + /* + * Zero bytes available from downChan because blocked. + * But nonzero bytes already copied, so total is a + * valid blocked short read. Return to caller. + */ - if ((error == EAGAIN) && (gotBytes > 0)) { break; } - *errorCodePtr = error; + /* + * Either downChan is not blocked (there's a real error). + * or it is and there are no bytes copied yet. In either + * case we want to pass the "error" along to the caller, + * either to report an error, or to signal to the caller + * that zero bytes are available because blocked. + */ + + *errorCodePtr = Tcl_GetErrno(); gotBytes = -1; break; } else if (read == 0) { + /* - * Check wether we hit on EOF in the underlying channel or not. If - * not differentiate between blocking and non-blocking modes. In - * non-blocking mode we ran temporarily out of data. Signal this - * to the caller via EWOULDBLOCK and error return (-1). In the - * other cases we simply return what we got and let the caller - * wait for more. On the other hand, if we got an EOF we have to - * convert and flush all waiting partial data. + * Zero returned from Tcl_ReadRaw() always indicates EOF + * on the down channel. */ - if (!Tcl_Eof(downChan)) { - if ((gotBytes == 0) && (dataPtr->flags & CHANNEL_ASYNC)) { - *errorCodePtr = EWOULDBLOCK; - gotBytes = -1; - } - break; - } - if (dataPtr->readIsFlushed) { /* * Already flushed, nothing to do anymore. |