Linux Today: Linux News On Internet Time.

More on LinuxToday

Kernel Compilation Project - Status update

Dec 12, 1998, 00:05 (0 Talkback[s])


Desktop-as-a-Service Designed for Any Cloud ? Nutanix Frame

Arjan van de Ven writes:

Kernel Compilation Project - Status update
This document tries to give a brief update about the Kernel Compilation Project.

In the last weeks, a lot of small but anoying bugs are fixed (bug thanks to Alan Cox for pointing me in the right direction) and a most fixes are in the AC-series. For some problems I have a solution but it is not "nice enough" to be in mainstream kernel.

Yesterday, I ran a series of compiles on 2.1.131ac7 + my fixes, and about 75% of the compiles was succesfull (BIG improvement, for example 2.1.125 compiled in about 3%). For the Kernel Compilation Project to work, this has to go up to (at least) 95%. Until that time, no distributed system is workable since a _LOT_ of the same bugs will be reported.

Remaining bugs

  • ISDN subsystem
    • The ISDN-subsystem calls icmp_send for a broken link; it shouldn't since TCP/IP is not always available (violation of OSI-layers)
    • Several HiSax-drivers depend on PPP for a CRC-table. This table is used very often, I think it should get it's own file
    • There is a nameclash between the act2000-isdn card and the hisax-drivers about a "cards" structure
    • ISDN_PPP depends on TCP/IP for VJ-header compression. Shouldn't since TCP/IP isn't always available (for example, for security-reasons)
    • The ISDN-subsytem doesn't compile if CONFIG_NET isn't set (Would be great for an old 386 to serve as answering machine)
  • PPP
    • PPP depends on TCP/IP for header-compression; same reasons as for ISDN
    • several drivers depend on the CRC-table of PPP; needs it's own file
    • "hfmodem_refclock_setscale" is called but neither I nor my compiler can find it
  • IrDA
    • Has several strange things (ac8 should fix; my machine is busy :-)
  • HDLCdrv
    • Fails if AX25 is compiled as a module but hdlcdrv isn't
  • linking-order
    • Several drivers depend on a linking order in the toplevel Makefile. Strange things going on here... Can give examples if someone feels like fixing these.
  • Radio-MIROPCM20
    • depends on the lowlevel-sound ACI-driver, but this one is asked later in the 'make config' process (yuck!)
  • Misc config
    • CONFIG_SDLA must be compiled as a Module
    • Promise DC4030 depends on CONFIG_BLK_DEV_IDEDISK
    • HIPPI depends on TCP/IP
Time to go fix :-)
Help appriciated! (.config-files available on request)

Arjan van de Ven