Skip to content

Fix Global Socket write() #171

@clauspruefer

Description

@clauspruefer

Current Result Processing

  • 1 Thread processes writes in serial order for 1 client socket

Review

This processing works and does not harm anything.

Problem(s)

  • Non blocking write() (how currently implemented) can in very seldom (heavy network processing) cause partial data processing (returns the processed byte count) if the underlaying send-buffer has only limited space
  • The current implementation takes only care of EAGAIN and not EWOULDBLOCK, both must be checked => bug 🐛!
  • On partial write(), write() call has to be repeated (but how? man pages do not help), this has to be tested, i assume only a repeated write() call without data (similar to sendfile() call)

Metadata

Metadata

Assignees

Labels

enhancementNew feature or request

Type

Projects

Status

No status

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions