On Monday 2024-11-11 19:34, Pablo Neira Ayuso wrote:
>On Mon, Nov 11, 2024 at 05:56:06PM +0100, Florian Westphal wrote:
>> egyszeregy@xxxxxxxxxxx <egyszeregy@xxxxxxxxxxx> wrote:
>> > rename net/ipv4/netfilter/{ipt_ECN.c => ipt_ECN_TARGET.c} (98%)
>> > rename net/netfilter/{xt_DSCP.c => xt_DSCP_TARGET.c} (98%)
>> > rename net/netfilter/{xt_HL.c => xt_HL_TARGET.c} (100%)
>> > rename net/netfilter/{xt_RATEEST.c => xt_RATEEST_TARGET.c} (99%)
>> > rename net/netfilter/{xt_TCPMSS.c => xt_TCPMSS_TARGET.c} (99%)
>>
>> No, please, if we have to do this, then lets merge the targets
>> (uppercase name) into the match (lowercase), i.e. most of the contents
>> of xt_DSCP.c go into xt_dscp.c.
>
>Agreed, please don't do this.
>
>We have seen people sending patches like this one for several years,
>this breaks stuff.
Because all those submissions renamed (e.g.) xt_DSCP.h to something
else.
It's kinda obvious that #include <xt_dscp.h> and <xt_DSCP.h>
must produce the same declarations as previously available.
Which seems doable with a layout like so:
xt_DSCP.h:
#include "xt_dscp_1.h"
xt_dscp.h:
#include "xt_dscp_1.h"
xt_dscp_1.h:
<the usual contents>
That way, xt_DSCP.h overwriting xt_dscp.h as a result of `make
install` (or some other file creation action) becomes
inconsequential since they have the same content under that model.
